Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
1
THE BROOKINGS INSTITUTION
AUTONOMOUS CARS:
SCIENCE, TECHNOLOGY, AND POLICY
AN EVENT FROM THE BROOKINGS INSTITUTION,
THE MAYS INNOVATION RESEARCH CENTER AT TEXAS A&M, AND THE TRANSPORTATION INSTITUTE AT
THE UNIVERSITY OF FLORIDA
Washington, D.C.
Thursday, July 25, 2019
Welcome:
ADAM LOONEY Joseph A. Pechman Senior Fellow and Director,
Center on Regulation and Markets The Brookings Institution
Introductions: LILY ELEFTERIADOU
Director UF Transportation Institute (UFTI) KOROK RAY
Director, Mays Innovation Research Center Texas A&M University
Keynote Speaker: KENNETH M. LEONARD
Director, Intelligent Transportation Systems Joint Program Office United States Department of Transportation
Engineering Presentation 1: Autonomous Vehicles, Traffic, and Humans:
LILY ELEFTERIADOU, Moderator Director UF Transportation Institute (UFTI)
SHERRILENE CLASSEN Professor and Chair, Department of Occupational Therapy
University of Florida CARL D. CRANE, III
Director, Center for Intelligent Machines and Robotics (CIMAR) University of Florida
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
2
PARTICIPANTS (CONT’D):
SANJAY RANKA Professor, Computer and Information Science and Engineering
University of Florida Engineering Presentation 2: Smart Infrastructure:
SWAMINATHAN GOPALSWAMY, Moderator Director, Connected Autonomous Safe Transportation Program
Texas A&M University SWAROOP DARBHA
Technical Director, Connected Autonomous Safe Transportation Program Texas A&M University
SIVAKUMAR RATHINAM Associate Professor, Department of Mechanical Engineering Texas A&M University
Keynote Speaker:
DEREK KAN Executive Associate Director, White House Office of Management and Budget Former Undersecretary for Policy, United States
Department of Transportation Panel 1: Policy Issues:
KOROK RAY, Moderator Director, Mays Innovation Research Center Texas A&M University
MARJORY BLUMENTHAL Senior Policy Researcher
The RAND Corporation RICK GEDDES
Founding Director Cornell University Program in Infrastructure Policy
CLIFF WINSTON Searle Freedom Trust Senior Fellow, Economic Studies The Brookings Institution
Panel 2: Public and Private Sector Collaboration:
TANYA SNYDER, Moderator Reporter Politico
JOYT CHADHA Senior Manager, NUMO Alliance
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
3
WRI Ross Center for Sustainable Cities
BARUCH FEIGENBAUM Assistant Director of Transportation Policy
Reason Foundation BETH KIGEL
Vice President, HNTB Director of Smart and Connected Solutions
TOM MADRECKI Vice President, Supply Chain and Logistics Grocer Manufacturers Association
* * * * *
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
4
P R O C E E D I N G S
MR. LOONEY: Thumbs up, okay. Well welcome to Brookings and thank
you very much for being here. I’m Adam Looney, I’m the Director of the Center on
Regulation and Markets here at Brookings. We’ve very excited to bring you today’s events
in collaboration with Texas A&M and the University of Florida. We’ll be gathered here today
for most of the duration of the day.
I wanted to remind you that today’s event is being live-streamed, and so
there will be people joining us from their homes and offices. You should all have a hard
copy that describes today’s program and the bios of all the participants. I just wanted to
emphasize that there are two panels this morning, there are two panels this afternoon.
There are two featured speakers. And in between I wanted to encourage you to join us for
lunch at the conclusion of the first two morning panels, and then for a reception that begins
immediately after the last panel.
And I want to begin this program with another introduction, which is an
introduction for our co-hosts from the University of Florida and Texas A&M. Dr. Lily
Elefteriadou, the Director of the University of Florida Transportation Ins titute, and the
Barbara Goldsby Professor of Civil and Costal Engineering. And Dr. Korok Ray, who is the
Director of the Mays Innovation Research Center at Texas A&M and a Professor of
Accounting at the Mays Business School.
Do you want to say a few words of instruction? Then come on up. And
thank you all for being here.
DR. ELEFTERIADOU: Good morning everyone. On behalf of the
University of Florida Transportation Institute I want to welcome you to this very exciting
workshop, Autonomous Vehicles, which is going to focus on technology, science, and policy.
As you know, this is a very exciting time in transportation, and we have
been partnering with industry, with academia, with the public sector, to develop and deploy
advanced transportation technologies. Florida is one of the largest states, the third largest in
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
5
the US, and has been for many years a pioneer in transportation infrastructure. And of
course no surprise, they’re now at the forefront of deploying advanced transportation
technologies.
So looking forward to the discussions today and the panels. And I want to
thank our collaborators, Korok and Adam, for putting together this event. I also want to
thank Anna for organizing the event. Thank you. Welcome.
MR. RAY: Howdy. It’s great to be here, thank you everyone for coming.
This is an exciting time for all of us. I feel like it was just yesterday when a few years ago I
was in my office coding in Python to gather some data for my research in economics. And
about an hour later I was on an abandoned airfield at Texas A&M with Shiva, and he was in
the backseat with me, also coding in Python, to navigate an autonomous vehicle. It really
showed me how technology has truly penetrated our lives. And in this conference I hope we
can talk about the technology and also understand all the opportunity it can bring.
So thank you all for coming. I want to especially give a shout out to John
Cannavo and Anna Dawson who did an excellent job coordinating this, we couldn’t have
done that without them. And we really look forward to all the exciting discussions today.
I believe this is going to be a little bit of a mix where we’ll have sc ience in
the morning and policy in the afternoon. I spent a decade of my life in DC, and I always felt
like there could be more science to some of the conversations, so this is our first attempt at
that.
So I hope you enjoy the experiment, and I look forward to meeting
everybody later today.
MR. LOONEY: I don’t know if I could pull off a howdy in the same way.
Welcome, Texas.
Well it’s my pleasure to welcome our first speaker, Ken Leonard. Mr.
Leonard is the Director of the US Department of Transportation Intelligent Transportations
Systems Joint Program Office. He spent more than 30 years in Federal government and in
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
6
the private sector, including at the Federal Motor Carrier Safety Administration and the
Federal Aviation Administration. He’s a recognized leader in the field of Intelligent
Transportation Systems, and he brings a unique perspect ive to what has to be accomplished
to do research, develop capabilities, and implement successful intelligent transportation
strategies.
I’m very excited that he’s here to open our program. And so please join me
in welcoming Mr. Leonard.
MR. LEONARD: Good morning. I am Ken Leonard, I am the Director of the
Intelligent Transportation Systems Joint Program Office. And for nearly seven years I’ve
had the great pleasure to work on an exciting portfolio including some of the most
fascinating topics in transportation science, technology, and policy. The JPO vantage point
allows us to look at intelligent transportation issues through a broader multi moto lens.
I’d like to start by thanking the Brookings Institution, Texas A&M, and the
University of Florida for sponsoring this event. I’m honored to speak on a topic that’s at the
top of USDOT’s innovation agenda.
The USDOT has worked to develop the tremendous promise of connected
and automated vehicles to solve some of transportation’s toughest challenges, to reduce or
eliminate the six million crashes, three million with injury, and almost 40,000 fatalities on the
nation’s roads each year.
We partner with a broad coalition of industry, academics, Federal, state, and
local agencies, safety advocacy groups, and others to support safe development, testing,
and deployment of intelligent transportation technologies. We’re working together to bring
connectivity to our roads, to create connected transportation ecosystem in which our cars,
our infrastructure, and our mobile devices communicate to yield quantifiable safety benefits
and an improved system efficiencies.
Why? Because connectivity is a problem worth solving. Connected vehicle
technologies have tremendous potential to reduce congestion, expand our individual
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
7
mobility, maximize system efficiency, all to improve the quality of our lives and, most
importantly, to save lives.
Innovation, and specifically ITS innovation, has the power to transform our
lives. We’re on the cusp of monumental changes. We see the maturing of past research in
connected vehicles as they move from demonstration to marketplace. And as we see a new
wave of emerging technologies such as automated vehicles, artificial intelligence, robotics,
fueled by the power of connectivity, data, advanced analytics, to machine learning.
We believe in the power of demonstrations and early deployments of
technologies to identify and overcome barriers to realizing ITS technology’s full potential.
Our demonstrations reduce risk and pave the way for widespread adoption.
Three of our largest connected vehicle pilot deployments in New York City,
Tampa, Florida, and across the State of Wyoming, are implementing applications tailored to
meet each of those region’s unique needs. The three locations are bringing together more
than 10,000 connected light vehicles, trucks, traffic signals, and pedestrian devices to their
communities.
Our pilot successfully demonstrated site to site equipment and application
inner operability to show that they can operate as designed anywhere in the country,
regardless of which competing manufacturers build them.
In addition to inner operability, we’ve worked out some of the cyber security
challenges. Since the inception of our connected vehicle program we understood the
potential vulnerability of information sharing among vehicles, and we incorporated the
concepts of privacy and security by design from the outset of the development of our
communications protocols for DSRC on 5.9. This is a critical design concept that is
overlooked at great peril in the rush to utilize spectrum for transportation purposes.
We also funded the development of a Security Credential Managements, or
SCMS, proof of concept demonstration which secured vehicle-to-vehicle and vehicle-to-
infrastructure communications. This system is now available via the commercial
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
8
marketplace to support deployment. In fact the US has over 80 ongoing and planned
deployments of connected vehicles. We know vehicle communications must be effective,
affordable, ubiquitous, and interoperable to meet transportation’s goals.
As communities begin to deploy connective vehicles, DOT remains
committed to protecting the 5.9 Ghz spectrum for transportation communications essential to
public safety today and in the future. Connected vehicles are just one example of our
research and investment in an advanced automated transportation eco system to improve
the quality of life for everyone and to achieve our vision to transform the way society moves.
We’re making progress not only in connected vehicles, but in automation as
well. In three successive years, from 2016, ’17, and ’18, we released substantive pol icy
documentation on automation. Our Federal Automated Vehicle Policy Guidance document
explored big picture questions around automated vehicles such as legislative authorities,
certification issues, and impacts.
Our Federal Automated Vehicle Policy 2.0 clarified the voluntary self-
assessment process and emphasized motor vehicle safety. The number of firms submitting
self-assessments has grown to 17, and it’s a number that changes almost week to week.
And while we’ve always remained committed to motor vehicle safety, we have expanded our
discussion of a more robust and multi-moto approach.
And released in 2018, AV 3.0 advances DOT’s commitment to supporting
the safe integration of automation into the broad multi-motoal surface transportation system.
It reiterates approaches to safety that were established in prior guidance, provides new
multi-motoal safety guidance, and outlines a process of working with the Department as the
new technologies evolve.
Preparing for the future of transportation, Automated Vehicles 3.0
establishes a clear Federal approach to shaping policy for automated vehicles based on the
following six principles.
First, we will always prioritize safety. USDOT will lead efforts to address
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
9
potential safety risks and advance the life-saving potential of automation and seek to
strengthen public confidence in these emerging technologies.
We will remain technology neutral, adopting flexible technology policies to
promote competition and innovation, and achieve safety, mobility, and economic goals. We
will modernize regulations. USDOT will modernize or eliminate outdated regulations that
unnecessarily impede the development of automated vehicles and that do not address
critical safety needs.
Fourth, we will encourage consistent regulatory and operational
environments. DOT will build consensus among state and local transportation agencies and
industry stakeholders on voluntary technical standards and advance policies to support the
integration of automated vehicles throughout the transportation system.
And we will proactively prepare for automation. We’ll provide guidance,
best practices, pilot programs, and other assistance to help our partners plan and invest.
But we won’t assume universal implementation of any particular approach.
And finally, we’ll work to protect and enhance the freedoms enjoyed by
Americans today. We embrace the freedom of the open road, and we envision automated
vehicles operating alongside conventional, manually driven vehicles and other road users.
These three policy documents are just the tip of the iceberg. And we’ve
taken many steps to advance the science and technology of connected and automated
vehicles with our research and deployment funding and convening activities and events.
And I’d like to share about a dozen of these activities with you.
In 2017 we hosted a round table on data for automated vehicle safety to
explore voluntary data exchanges to accelerate the safe deployment of automated vehicles.
As a result of that round table we’ve created a minimum viable product for automated
vehicle works on data exchange with a number of states, and multi-state pilots are
anticipated soon.
We hosted over 1,000 stakeholders at a public listening sessions summit on
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
10
automated vehicle policy in March of 2018 to seek input and broadly engage the public.
USDOT modes have published requests for information from stakeholders and we’ve
analyzed thousands of comments that we’ve received.
Several modes have engaged with their stakeholders through face-to-face
national dialogues and public listening sessions, some just as recently as last week at ABS
in Florida. And we’re assessing national issues, developing technical guidance, supporting
research, adopting programs, policies, and creating a national community around
automation.
One of our most important partners in solving automation issues is
Congress. And in response to the Omnibus 2018 Appropriations Act, we developed a plan
that communicated our management and research demonstration plans, as well as
coordinated across the Department on over $100 million in additional funding going into
automation.
As part of the Omnibus, in December of ’18 the Department published a
NOFO, a Notice of Funding Opportunity, for ADS demonstration grants. These grants will
provide up to $60 million in funding for projects that test the safe integration of ADS on our
nation’s roadways. We received a lot of interest in the grant funding opportunity, and the
application closed in March. We expect to announce awardees soon.
We’re also partnering with the Department of Labor, the Department of
Commerce, and the Department of Health and Human Services to conduct comprehensive
analysis of workforce impacts and operational safety issues for commercial drivers. They’re
introduced through automation technology.
We held an event in March to gather stakeholder input and we’re in the
process of developing a study and will publish a report this summer. Our goal is to provide
information to policymakers and the public to help the nation prepare so that we can all
benefit from the introduction of new technologies such as automated trucks.
NITSA, FMCSA have issued advanced notices of proposed rulemakings on
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
11
the removal of unnecessary regulatory barriers to the safe introduction of ADS vehicles in
the United States. Both agencies are seeking comments at this stage to ensure that all
potential approaches are fully considered. As the agencies move forward with these
regulatory actions, we encourage you to submit comments to the Federal Register Dockets,
comments on both AMPRMs are due July 29th, so just around the corner.
The Department has also sought public comment on proposed exemptions
to Federal standards and how the public can be protected as new transportation
technologies emerge.
In March NITSA issued two requests for comment in response to petitions
for the temporary exemption of the Federal Motor Vehicle Safety Standards, FMVSSs,
submitted by General Motors and NURO respectively. The comment period closed in May,
and NITSA’s reviewing the comments that they’ve received.
Federal Highways has developed the Carma Platform to encourage
collaboration between transportation systems management and operations strategies and
automated driving technology, focused on how infrastructure can move traffic more
efficiently.
And three awards were made for truck platooning projects to assess in-
service truck platoons delivering commercial goods via fleet operators on their common
delivery routes, to understand how truck platoons will operate in a realistic operational
environment. Results will inform state and local decision makers that develop truck
platooning regulations.
And lastly, but certainly not least important, in April and June of this year we
held two public sessions on the 5.9 Ghz safety band. The safety band is critical to ensuring
connectivity that enables cooperation through secure, uninterrupted, below agency, inner
operable, proven communications. Protecting the 5.9 Ghz band radio spectrum, the safety
band, is of critical importance to the Department. It’s uniquely capable of supporting safety
applications that prevents or significantly reduces the severity of vehicle crashes in a
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
12
manner that’s not available through existing vehicle technologies.
The safety band is already used by state and local transportation
departments for vehicle to vehicle and pedestrian collision avoidance as well as emergency
vehicle priority. The safety band also enables applications for red light warnings, reduced
speed warnings, curve speed warnings, spot weather impact warnings, and other safety
critical applications.
As technology advances inner operability remains central to enabling
universal vehicle to everything capability and benefits. USDOT must ensure that the safety
band is protected for traffic safety. The Department is technology neutral, but we’re not
technology agnostic. We believe that technology must perform. It must deliver on the
promise of safety, safety is our highest priority, and transportation technology using the 5.9
safety band must demonstrably achieve it.
At the same time we encourage the automotive industry, wireless
technology companies, and other innovators to continue developing multiple technologies
that leverage the 5.9 Ghz band of spectrum for transportation safety benefits.
So as you can see, we’ve been pretty busy. USDOT is moving forward with
both connected and automated vehicles, we’re committed to engaging stakeholders to
identify and solve scientific technical and policy issues, and usher in a new era of
transportation. The Department will work with partners to integrate connectivity and
automation across the transportation system, enhancing mobility for seniors, people with
disabilities, facilitating freight movement, and increasing productivity . The DOT’s actions
represent just a step in the right direction. We know that it’s a long road, but one that we
must take together. Connected vehicle technology is proven and in deployment.
Despite its promise and the progress that’s been made, automated vehicle
technology is still in its early stages of development. The public has concerns about the
safety and security of this new technology. These concerns must be addressed because
without public acceptance we know automated technology will never reach its full potential.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
13
Ensuring that we establish common methods to evaluate the safe performance of emerging
ADS equipped vehicle technologies is critical. DOT and industry are working to assure the
safety of ADS equipped vehicles. The Department supports automation in a safe and
secure environment where innovation can thrive and the American public can be excited and
confident about the future of transportation. The promise and safety of automated vehicles
is only possible if we all work together.
So in conclusion, and in advance of your discussions today, let me thank
you again for your work to advance smart transportation. So thank you.
DR. ELEFTERIADOU: Good morning again. This first part we will focus on
the technical and scientific aspects of autonomous and connected vehicle technologies.
First I’ll start with an overview of those technologies and their deployment.
My background, just to give you an idea, is on traffic engineering, my expert
in traffic flow theory and civil engineering.
Dr. Carl Crane is a Professor of Mechanical Engineering at UF, he’s been
working on autonomous vehicles for several decades. He’s developed both surface and
water vehicles, autonomous vehicles, and he’ll discuss the technology and sensors for
autonomous vehicles, along with deployment challenges.
Next Dr. Sherrilene Classen is a Professor and Chair of Occupational
Therapy and she’s an expert on fitness to drive issues and driving simulators. And she’ll
discuss the interactions of technologies with humans, and particularly perceptions by older
drivers.
Dr. Sanjay Ranka is a Professor of Computer Information Science and
Engineering. He’s an expert in computing and machine learning, and he’ll discuss
transportation related data management applications that are under way.
First let’s start with the status of the transportation system in the US. And
Ken mentioned the more than 40,000 fatalities in the US this past year. This graph shows
how the number of fatalities have changed since 1975, along with the fatality rate per 100
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
14
million VMP. As you can see, both numbers have been going down. And even though
vehicles and the infrastructure have generally become safer, the vast majority of these
crashes, 94 percent, are due to human error. So the promise of autonomies to reduce these
substantially or even eliminate them.
You have seen various tests by Google, Uber, in the news, and now there
are more than 1400 vehicles being tested by more than 80 companies across the US. We
see more and more shuttles, such as the one shown here on the screen, which is the one in
Gainesville.
So as you can see, the vehicle is going really slowly. It’s really frustrating if
you’re behind it but it is because it’s mapping the route. And during that time, I don’t k now if
you caught that, but there was a bicyclist zooming by, and the vehicle just stopped.
Wouldn’t it be nice if all our vehicles could do that? And this by the way, was not staged,
this is something that just happened while the vehicle was mapping the route.
This shuttle is going to be part of the original transit service in Gainesville.
We’re waiting for approval from NITSA, the vehicles themselves are in Gainesville. But we
hope this will give us great information about interactions with the public, how people might
use autonomous cells.
Now let me take you back to when vehicles were first invented. That’s what
they looked like. So we didn’t get to where we are overnight, it took about more than 100
years to get to the transportation system of today. And at that time acceptance of vehicles
was very far from universal. There were significant objection, significant safety issues, no
rules of the road, no stop signs, no traffic lights. And at the time of course horses were
prevalent with problems, manure, mistreatment of the animal, and so on. So mixing the two
was a challenge. The horses were getting scared of the cars.
So this little invention was introduced. It’s a horse head so that horses
wouldn’t be spooked. Of course that didn’t work. But in some ways we face similar issues
today with humans accepting autonomy.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
15
How will conventional vehicles and drivers operate their own autonomous
vehicles? Will they cut them off because they can, because they can trust that an
autonomous vehicle will stop? How will the autonomous vehicle interact with pedestrians
and bicycles and scooters, which are very prevalent on the UF campus?
So the concept on this slide is kind of similar to the horse head. Now we
want to make the vehicle look and act like a human driver so the pedestrians and other
human drivers can understand what the intent of the vehicle is. So if the vehicle is
approaching, are they going to stop for me to cross if I’m a pedestrian? In the autonomous
vehicle logic also wants to know is the pedestrian standing on the side of the road getting
ready to cross, or are they just waiting for something else?
So in my field, in traffic engineering, we use this graph in introductory
transportation engineering courses. And it shows the three major elements of the
transportation system, the vehicle, and its characteristics, the driver, the traveler, and then
the infrastructure and the environment. So each of these affects and is affected by the other
elements, and all together contribute to the performance of the transportation system.
So the speeds that we observe are a product of all these. The capacities
are a product of all these. So as an example, when you have a heavy truck, it’s traveling
much slower when you have a steep upgrade. So the combination of the vehicle
characteristics and the infrastructure result in particular performance. Each of us as
humans, we react differently to the different designs. Some of us are more cautious, some
are more aggressive, some of us we want to keep longer distances, shorter distances. So
this variation that we have from human drivers also affect the speed, the travel times and the
safety.
So how do these fit within this system? How will different travelers use and
interact with AVs, with autonomous vehicles? How will different vehicles from different
manufacturers operate under different highway designs? So when we’re introducing
autonomous vehicles we have to consider all three elements. We have to consider the
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
16
variabilities in driver behavior, both in using a particular autonomous vehicle and driving, as
well as interacting with a pedestrian or as a bicyclist.
So here is a graphic, this type is commonly used to describe autonomous
vehicle operation. So an autonomous vehicle maps the terrain. It has several types of
sensors by which they’re able to look at obstacles and navigate through the highway
network. Now these are detected with the bigger sensors, then the vehicle’s logic decides
how it’s going to move, whether to stop, and whether to change lanes.
But imagine this scenario on this slide where you have an autonomous
vehicle which approaches a crosswalk, and there’s a pedestrian already on the crosswalk,
there’s a bicyclist further ahead. And this can be seen by the sensors. But there’s another
pedestrian standing there. And its view may be somewhat obstructed by the sensors on the
vehicle. So can the vehicle detect the pedestrian, first of all? And even if it can detect it,
can they determine the intent, whether the pedestrian is getting ready to cross?
When we have a human driver, the human driver can communicate that
non-verbally with the pedestrian, determine the intent, allow them to motion them to pass,
and so on. And this becomes a very difficult problem, especially considering the viability in
driver behavior. So different individuals will behave very differently.
Now imagine this environment. Freeway congestion, and then an
autonomous vehicle operating in it. In this environment you’d have very frequent lane
changing as vehicles are trying to get to the correct lane for their destination. And so when
you have an autonomous vehicle under this environment, the autonomous vehicle has to
recognize the intent of other vehicles changing lanes.
Second, you have a distance that the autonomous vehicle needs to keep
with the vehicle ahead to maximize its safety. If that distance becomes too long then the
space that this que occupies with conventional vehicles becomes much longer. If it’s too
short it may cause rear end crashes. So the vast majority of autonomous vehicle crashes
that have been recorded so far are due to human error or of a conventional vehicle crashing
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
17
into an autonomous vehicle, because they don’t react quickly enough. The human driver
doesn’t react as quickly as the autonomous vehicle.
So autonomous vehicles do have the potential to significantly improve
safety and mobility, but there are risks in their deployment. And those deployments may be
derailed if we start increasing delays and crashes. So the wrong type of deployment, when
the technology’s not ready or well thought out, will backfire.
Let me show you now an example of how we’re approaching this problem to
avoid these pitfalls for deployment of autonomous connected vehicles. We have a
partnership at the University of Florida with the Florida Department of Transportation and the
City of Gainesville to build the I Street, which is a real world test bed. And its objective is to
test, deploy, evaluate, several different advanced transportation technologies.
Gainesville and the UF campus where this is deployed are ideal for this type
of testing because there is significant pedestrian, bicyclist traffic, there’s scooters. We have
an extensive transit network. We have relatively low speeds, and good weather. Also
Florida is very welcoming to autonomous and connected vehicles. So the Florida
Department of Transportation has funded the deployment of communication devices for
connectivity along several corridors, including corridors around the UF campus. And those
allow vehicles that are instrumented to communicate with infrastructure.
So what I’m showing in this slide is a system we’re developing for optimizing
signal controls, so the operation of traffic lights and the way that the green is allocated to the
traffic approaching the intersection while at the same time advising arriving vehicles that are
able to receive this information what their optimal path is, or their trajectory.
So the autonomous vehicles approaching the intersect ion, as soon as they
reach the communication range, they receive very specific information on the path that they
should follow, and they can execute it. We’ve been doing experiments now for a couple of
years and we’ve developed the hardware and software to be able to do this.
With connected vehicles, which are also part of the mix, if they have the
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
18
ability for connectivity then they can also receive information on optimal speed. And right
now we’re working on an app where we are able to have even convent ional vehicles with no
connectivity be able to receive information on optimal paths, advisory speeds, in essence.
This picture shows how we use radar and other sensors. This is at an
intersection on the UF campus Stadium Road and Gale Lemerand. You see our football
stadium there on the left. So that intersection is instrumented so that when the vehicle is
connected it can receive an optimal path, which is the green point there.
So when we first detect a conventional vehicle we estimate its projected
path. We optimize signal control. And then as the vehicle moves we reevaluate the path
and adjust the signalization if the vehicle was slower or faster than what we predicted.
So with initial simulation experiments and for a fully autonomous simulation
environment, but with communication with the infrastructure, our experience shows that we
can reduce delay by 20 to 70 percent, depending on the demand. Of course that number
goes down when you have mixed traffic and lower market penetrations. We start to see a
significant difference when we get to about 20 percent market penetration.
So a system like this cannot function if it doesn’t have communication with
the infrastructure. And to be able to take advantage of such systems we have to have
collaboration on both the public sector and industry, academia. And so I Street is
approaching this as a collaboration with our partners, Eftason and the City of Gainesville.
Because if we don’t get this right, if we don’t make a difference, then congestion wil l get
worse and deployments for autonomous vehicles will backfire.
So this map here on the right shows our ongoing deployments around
Gainesville for I Street. So I invite you to check out our work on I Street. You will hear some
more with the next three presentations, and we look forward to working with you to improve
mobility and safety.
Thank you. And with that I’m going to get Dr. Crane here for the next
presentation.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
19
DR. CRANE: Well, thank you very much. We call her Dr. Lily, so she’s a
fantastic leader of our efforts down in Florida.
So I was asked to try to do some -- I was watching that clock there, a 15-
minute overview of autonomous technologies to a group of experts in a wide variety of topic
areas, which is usually a recipe for disaster. So here goes.
A little bit of the over -- I’m glad you started with the old cars. I hadn’t seen
the horse head, I thought that was great. Background. I’ve been in Florida a long time. I’d
been in the Army before I got to Florida, and like in the Army, every three years you transfer
and you’re just -- so my three years got up, I was like great, I’ve been there now for like 35
years, so I guess it’s not going to happen.
But over that time we’ve done a lot of autonomous vehicle work for the Air
Force, for different projects that they had. And then in the mid-2002s DARPA had their
Grand Challenges and the Urban Challenge. I was going to say we participated in all three.
The Grand Challenge, if you’re not familiar, was they gave along a desert
route latitude/longitude of a whole bunch of points so about 150 mile long path, and whoever
could do it the fastest would get, well they upped it to a $2 million prize. And the first year,
shoot, five vehicles got about five miles, and five got about a mile, and five self-destructed at
the starting gate. And they were the best 15 out of 43. We were in the middle group, we got
about a mile. My goal is make sure we get out of sight of the starting gate, and then
whatever happens, happens.
But the second year though definitely, it was about a year and half later they
redid it. And 43 teams, they put 23 in the race. Five completed, and the fastest was
Stanford, and they got the $2 million prize and kind of led off to Google, Waymo, and all that.
We got about 14 miles and our GPS went out. Well our GPS, which is supposed to
broadcast corrections to a few centimeters level accuracy had an eight meter error in GPS,
and the entire vehicle was a single point failure vehicle. So I can’t say that I was
disappointed, that’s not the right word. But that was actually our high water mark, I think.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
20
And it was really exciting to see that happen. So now I used up all that time on the first
slide.
What I want to do today is just show, you know, sort of a basic overview of
what’s going on with autonomous vehicles. And I don’t have to spend a lot of time on this
chart. But what really jumps at me is that orange word “deaths.” And when you see that
and it’s next to that blue line and it’s going way up high, you say “Oh, my goodness, this is
terrible.” But it turns out that that’s actually vehicle miles traveled. And one of the other
increasing one is population. And if you go to the very bottom line about deaths per billions
of vehicle miles traveled, there’s been real improvements. And I think a lot of the safety
factors that you had mentioned. I’m going to get a new car this fall and I’m going to get
everything I can on it because it’s amazing what they’ve got up there.
So there has been big improvements, but there can be better. So if I look at
the basic problem for, you know, one autonomous vehicle, basically all you have is your
instantaneous sensor data, and you have whatever operative or map data that you have
ahead of time, and that’s all you have. And from that, again, in every instance you have to
get your throttle, steering, braking, whatever you want to get to come out of it.
And for the instantaneous sensor data, you know, one is localization. Can
you get some idea, you know, a typical GPS might be, you know, for meter accuracy, it’s not
too bad. You can do better than that with some other technologies. Identify static objects,
the terrain, where is it smooth, where are the curves, and then lane markings. But the
moving objects are the big one, the other vehicles and pedestrians and things like that, how
do we do it.
The main sensors that are out there, I think I’ve said the accelerator pedal, I
gotta go a little faster. The main sensors that are out there, LIDAR, you know, is very
popular, you shoot the laser and you get this point cloud, you get this point data. And the
advantage is 80 meters is easy. We’ve got one sensor that claims I saved 300 meters that
we used, oh, the Urban Challenge, that’s the city street DARPA competition, I forgot to talk
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
21
about.
But you might miss the target. The beams, you know, it’s not continuous,
reflections due to rain or things like that, or certain materials don’t reflect the material, you
know, the beam that well.
Our vision is great because it’s passive, you’re not sending any signal out
there, but again, lighting and weather conditions and something as simple as the stripe in
the road going under a shadow of a tree can cause some issues sometimes.
Ultrasonics, I know we started way back with some of our Air Force work,
we were moving very slow, it was construction equipment, and ultrasonics is the poor
person’s sensor. They are inexpensive, but short range, and all you know is in this cone of
area there’s something out there but you’re not sure where.
And radar is, you know, probably one of the best as far as bad weather and
things, so that you can get it out there. But again, it’s not pinpoint sort of like a laser beam
as far as locating where something is out there.
This is our Urban Challenge vehicle, so that was a competition DARPA did
out at an Air Force Base that was closed in Victorville, California. You had to drive on city
streets. Again there was a $2 million prize, but it was a bit more subjective this time.
Instead of who did the 150 miles the fastest, it was clear who the winner is. This time, so
first place was Carnegie Mellon got $2 million, Stanford got $1 million second place, Virginia
Tech third, got half a million dollar prize. If I was to ride in one of the vehicles, maybe I
shouldn’t say which one I would want to ride in.
So this is our vehicle, it’s a Toyota Highlander that we did. We have a lot of
single beam radars down at bumper level, you see Sick, it’s a German sensor, pronounce it
Sick. I don’t know. Sends out a single beam, so on the bumper level there’s sort of a plain
of beams that are out there. And if you look at this top-down view, there’s an L shape right
in the middle, has a number 36 next to it. That’s what from the top view, but that’s another
vehicle. So you’ve got to interpret that as a vehicle. Some of the other objects, you see how
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
22
we grouped them together anyway, our buildings and things, the vehicle itself is that
rectangle, it’s not labeled, in the middle.
And then as you move forward, this is some of the technology where the
prices have come down now. This is a Velodyne 16 meter LIDAR that we use quite
regularly. And you see how it’s angled on the left-hand picture so that it swings 16 beams
360 degrees. On the left of the blue dot thing, all that mess up there are the trees because
the back side of the LIDAR is shooting up there. But on the right-hand side we see the 16
beams as they’re sweeping across the road. So again, you don’t get continuous total data
but you get all these points as indicated by those blue dots.
This is a 16 beam LIDAR as we put on our vehicles when we’re driving
around campus, so again, you get a lot of information. The problem becomes how to
interpret that fast. And the 64 beam LIDARs are very popular. When these first came out
though I think the 64 matched to the price, I think it was $64,000. Yeah. But you get a lot of
data. And you can say “Why is that important?” Well, we’re going to get to that.
What about the operary information? All right. We’re going to have maps,
so how are we going represent those maps. Are they going to be point clouds, are they just
going to be -- well let’s just show the options.
This is the map that we had for the DARPA Urban Challenge. It was a
texted file. And if you look down the list here, you can see about halfway down on the left it
says “Number of lanes, two.” And then two lines below that it says “Lane 1.1, number of
weight points 4.” And down at the bottom of the left-hand column you can see four pairs of
numbers. Those are latitude/longitude of certain points. And then it says “End of Lane 1.”
And then in the right-hand column, Lane 1.2, so it’s Segment 1 lane 2, has six weight points.
And you can see the lat/long. And the fifth line it says “Exit, 1.2.4, and then 3.1.1. That
means that you can exit or you can transition from Segment 1 lane to .4 to Segment 3 lane
1.1.
So that was our world, that’s all we had. And from that then, you know, we
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
23
planned routes to -- oh, we were given a mission file so you go points 8, 11, 20, you know, in
a certain order. So you had to plan your route with that information.
So do you have a map like that, or do you give entire point clouds? So this
was us driving down one of the roadways in Gainesville with a 16 beam LIDAR and just one
single path. And you can even see the wires in the power lines. And it’s a lot of information,
but it’s all available.
Or do you take all that information and try to condense it down so you can
make a Boxall representation so it ’s sort of line Minecraft representation. So on the left you
see all these boxes. And on the right we did some classification. This is something for the
Air Force, we were looking for tree trunks and not people. And the purple we identified as
tree trunks, except for the one image.
So if you had instantaneous sensor data and you have some type of map
data, how do you decide what to do? I’m not going to dwell on this slide, but one of the
biggest things we did on our Urban Challenge vehicle was how do you decide what behavior
to do. Again, it was to go from that sensor data, which is the blue thing on the bottom, to
throttle breaking. I’m going to zip through that one.
Okay. What can you get today? And I have to apologize to some of the
other vendors, this is a Tesla video, but General Motors has some great stuff, a lot of the
people have great stuff. So if you were to go to the showroom today and buy something,
you can basically be on like an interstate and do lane changes, stay in your lane, stay
behind a vehicle. Some of it’s kind of cute, you know, tells it to pull out of the garage and
come get ready to go. And, you know, automatic parking, you know, it’s been out there quite
a while. And the last one is emergency braking. And those are, they’re really good things,
and I’d love to have that in my car, but it’s not an autonomous car.
So again, kind of rushed through in 15 minutes, and you go from what’s
available to buy today, go to some of the esthetical topics, maybe some of the things that
keep us up at night. And then go back to the LIDAR. So the problem with the LIDAR is how
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
24
do I take, for example, this point cloud. So on the left are all the points that were hit, I
believe there are 16 beams here. Somebody was about two and a half meters away, you
get a lot of data. And you can see on the right-hand side, you can hardly see the dots at 17
meters. And 17 meters isn’t that far but you don’t get that many hits. So how do I identify
that as a person?
So this is from a TED talk from Chris Urmson from Awaymo. One of the
approach was they generated all these data of actual people and have stored them, and
then you compare, you know, the point cloud that you have with the models you have and
identify.
And then this is one, I just wanted to show a case. I don’t want to say things
that keep us up at night, but let’s see if can I get this to play. Play, play, play. Okay, so let’s
play. There’s an intersection indicated by those orange lines. And the mouse was wiggling
by a person there. And this is the raw data, and we classified now. There’s the person, and
the person’s about to walk across the road. And you can see we’re getting great point hits.
The sensor is located at the intersection of the blue/green/red lines. And as they walk
across, look how it disappeared. Now if you look at the raw data there’s still data there, bu t
our algorithm said there wasn’t enough points to classify that was a person. So that’s a
problem. And it’s not a problem that’s going to go away. Now on Tuesday the student ran
in, hey I got a new video, I’ve got that case solved. And I didn’t bother to put it in the
presentation because we’ll have a new problem tomorrow anyway.
But those are the kind of things that cause, you know, trouble to get out of
there.
So we’re going to go to issues. Like I say, it has to work 100 percent of the
time. Nothing works 100 percent of the time, but that’s not the real issue. When the system
realizes it’s having trouble it has to be able to turn over to the human, you know, in enough
time, things like that. So the whole system has to work 100 percent of the time, let me put it
that way.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
25
I worry about, and maybe this is what I should have said, the stuff I worry
about, sensor of performance, important environmental conditions, false positives/negatives.
This is one I think Dr. Classen’s going to talk about, is things like this. If I’m in the vehicle,
do I want the vehicle to explain to me what it’s doing? I’d feel better if I see this map and I
see little boxes around the other vehicles, that it’s identified them and I’ll feel better. I’m not
grabbing the wheel though. And then the communication issue, I think that’s going to be a
really big plus vehicle to vehicle. Legal/ethical issues.
So what I’m going to end on is I got two videos. And I think this is from
Vietnam. Intersection of the present. Maybe it’s efficient, they’re all moving, nobody is
stuck. So we go from that. And I found this, it’s at low resolution and the reference, it was
from France. Here we go. They put a lot of time into this I believe. And it goes on for a little
bit longer. So, you know, that’s definitely not what Gainesville looks like today.
So thank you for your patients. Again, I tried to give a shorter, quick,
shotgun overview of the technologies that are out there. And we’re going to turn over to Dr.
Classen now for the next.
DR. CLASSEN: So after all of that I feel very compelled to represent and to
present the human perspective to be able to see how folks are experiencing all of these
exciting growth and challenging times as we are really standing in the transportation
revolution.
I’m an occupational therapist, so what ODIS are doing is we are interested
in the person, the vehicle, and the occupation. Or, those meaningful tasks that each one of
us have to do to be able to survive, to strive, to thrive, to live meaningful lives, and of course
to drive.
So from that perspective I’m going to talk about a study that we’ve
conducted at the University of Florida to look at older adults’ perceptions of autonomous
vehicles. So I’m going to use the scientific outline to be able to frame my presentation. So
it’s going to be a true scientific presentation.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
26
And a little bit of background in terms of older drivers, or older adults. So
we know that the demographic shift in America is happening. And about 20 percent of the
US population is over the age of 65. And that is to increase in the next 10 years, so that we
will have one out of four drivers will be over the age of 65.
Despite the fact that older drivers have issues with medical issues, declines
as a result of the normal aging process in their cognitive, their visual, their motor, and their
sensory functions, as well as being on complex medication regimes. And that causes that
they are at a higher risk for crash related injuries and fatalities. Despite that, driving is still
the preferred mode of transportation among this population. So autonomous vehicles might
hold excellent opportunities for these drivers to offset some of the health as well as the
safety issues.
Recent perceptions were measured by survey only where older adults were
asked about autonomous vehicles, survey only. So we thought well, if we can use the
surveys, but if we can also expose older adults to some of these autonomous vehicle
technologies, what will we see? And the expectation that we had going into the study was
that their attitudes, their perceptions, their values, and even maybe their early adoption
thoughts may improve as a result of exposing them to some of these technologies.
So what we did is we exposed them to a driving simulator that was
programmed at the SAE Level 4. And we also exposed them to the autonomous shuttle, the
one that Dr. Lily showed in the previous presentation. So obviously because we are working
with human subjects, we had to go through ethics approval and received a full IRB approval,
that’s the Institutional Review Board approval from the University of Florida. And we
developed a repeated measures crossover design, and I’ll say a little bit more about that in a
second.
We recruited the participants from the northcentral Florida area, but with a
concentration in Gainesville. And we included folks if they were at or older than 65 years of
age and if they had valid driver’s licenses. And the exclusion criteria pertained to their
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
27
performance on the Montreal Assessment, cognitive assessment and using the indicators of
that particular assessment to understand cognition and cognitive declines.
We had a variety of measures that we used. And I’m not going to go over
all of them, but on the left-hand side you can see some of them pertain to chemical tastes,
looking at safe shifting as well as user perception surveys. And then also the driving habit s
questionnaire which basically calculates or describes all the drivers’ habits. And the life
space questionnaire to help us to understand how older people are moving in their home,
outside of their home in the neighborhoods, or do they even cross state l ines as well.
So, let me just go back to this one. On the right-hand side you see the RTI
high fidelity driving simulator. So in this simulator it’s got a full car cab, and the older driver
would sit behind the wheel with a research assistance in the passenger seat. These life-size
images that’s being projected on the actual screen. And, folks, the experiences are very
immersive experience with very high what we call physical and emotional fidelity. Meaning
that the believability of this situation is really high.
On the bottom is the transitive easy mile, easy 10 autonomous shuttle that’s
also been programmed at the level for the SAE standards, and that’s the shuttle that we’ve
used. In this particular shuttle we had six older adults at a time in the shuttle, and we had a
research assistant as well as the safety operator accompanying the older drivers . And
because of not having the actual NITSA waiver to run the shuttle on the public roads we
conducted this study in a bus depot in Gainesville.
So after folks enrolled in the study, they completed all of the base line
measurements that I’ve mentioned in the previous slide in visit one. And then they were
randomly allocated to Group One and Group Two because of the crossover design.
So for folks in Group One, they were first exposed to the simulator and then
they were exposed to the shuttle. And after each one of these exposures we actually used
the AV User Perception Survey to look at their perceptions prior to exposure, exposure after
the simulator, and exposure after the shuttle, to see if there’s any changes.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
28
And in Group Two we flipped the order of the exposure, and they were
exposed to the shuttle first and then the driving simulator.
So we collected the data by trained research assistants, and we entered
that in a red cap system that adheres to all of the privacy and the confidentiality policies at
the University of Florida. And our analysis was conducted with statistical software, SPS
version 25. And today I’m going to give you a preliminary look into the data that we found.
We haven’t completed the study yet, we are enrolling 105 older adults, and this is 69 older
adults that’s completed the study. So please see this as preliminary data. And of course we
cannot make any causal conclusions as a result of that.
I just want to say something about the user perception survey. When we
started the study there were no user perception surveys that actually asked the questions
that we wanted folks to answer so we had to design our own user perception survey. So this
is still undergoing psychometric testing and so it’s in development as we ll.
But some of the things that I would like to highlight is that we basically have
28 questions on the table over here -- the pointer is not working, but you can see that there’s
a total of 32 questions, 28 of them are all questions pertaining to perceptions, values,
beliefs, and experiences with autonomous vehicle technology. And then we ask four open-
ended questions such as the disadvantages or the advantages of autonomous vehicles.
So all of these items derive from basically the dimensions, which is in the
top left-hand corner. So we look at questions pertaining to trust, safety, control, experience,
etcetera. And all of those dimensions are basically conceptually imbedded in some of the
existing models that’s at the bottom there, there are seven existing models. So as I say,
psychometric testing is continuing with this particular survey. But this is the survey that we
used for the study.
So if we are looking at the results. So this is some of the demographics of
our participants. So for shuttle first exposure we had 20 women and 18 men. And the
women were a little bit younger than the men in comparison. So 73 was the mean age, 77
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
29
the mean age of the men. And that pattern pretty much held for folks who were exposed to
the simulator first as well.
If we look at the gender distribution we had so far 29 men, that’s 42 percent
of the sample, and 58 percent female. And that’s a pretty good representation of the current
demographics for gender, according to the Florida census statistics. In Florida it’s 45
percent male and 55 percent female. If we look at racial distribution, so we had 87 percent
of the folks that self-classified as Caucasian or white, nine percent classified as black, and
four percent classified as other.
This is an interesting slide in terms of the actual education. So Gainesville
being a college town, and much of our recruitment occurred amongst the older adults in
Gainesville, you can see that 73 percent of our participants had either Bachelor’s degrees,
Master’s degrees, or Doctorate degrees. So that is totally of course skewed in terms of the
general population. We also had 23 percent of our participants having a technical school,
some college, or an Associate degree, and four percent of our participants had a high school
diploma or a GED. So a little bit skewed in terms of education. So we worked with a very
highly educated group in the study so far.
All right. So I’m going to show you the results of the entire group before the
actual exposure to any of the technology, and then exposure to the shuttle and the
simulator.
So a brief orientation here is that on the YXs you will see the user
perceptions. And we measured that with what we call a visual analog scale. And so that’s a
scale that is numbered from 1 to 10, it’s exactly 100 millimeters, and we use that as numeric
data to be able to measure the perceptions of older adults. And on the X axis we have the
baseline condition, the shuttle condition, as well as the simulator condition.
So the variables on the right-hand side pertains to the social influence, the
intention to use, and the cost. And all of these, if the graphs is going up it shows that older
adults are more favorable to each one of those conditions.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
30
So you can see that was the case for being exposed in the shuttle as well
as in the simulator, although there was a dip when folks were exposed to the simulator. The
outlier here is the intention to use, where the older adults were really more positive after
they’d been exposed to the simulator, and the intention of course is to use autonomous
vehicle technology.
For the actual next three variables, this is pertaining to how media influence
people as well as their trust of autonomous vehicle technology and the controls. Do they
want to take control over of the vehicle or would they rely on autonomous vehicle
technology. You can see that the graphs are pretty much telling us the same picture. At
baseline, you know, they are lower. Once they get exposed to the autonomous shuttle,
sharp increases, and then to the simulator we see that the graph is decreasing again.
And then our last four variables pertains to authority, excuse me, that’s
pertaining to trusting the National Highway Traffic Safety Administration’s safety measures,
perceived usefulness, perceived ease of use, as well as safety under each one of those
cases. Again you see that folks felt much more positive to these four variables after they’d
been exposed to the shuttle versus being exposed to the simulator, but still the simulator
had an effect.
So if we then looked at what happened in the groups. So if we look at
Group One, you will see that they were first exposed to the simulator and then to the shuttle.
Now I don’t want to be saying that we can make causal inferences, but we did look at
statistical significance, and that’s outlined in the red font over there. And we also looked at
effect sizes to tell us how big the effect is.
And what we learned from this first table on the left-hand side, is that the
older adults’ trust increased quite a bit after they’d been exposed to the shuttle last, as well
as their perceptions of safety after they’ve been exposed to the shuttle last .
And then for Group Two where they’ve been exposed to the shuttle first and
then the simulator, we see that the intention to use increased. So we think it’s because in
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
31
the simulator they were sitting behind the steering wheel, it felt like they were an operator of
the simulator, where as in the shuttle they were more passive as a passenger.
Okay. So my take home messages, you know, overall analysis, exposure to
AV technology positively effects older adults perceptions to this emerging technology.
Perceived safety shows the highest percentage points of increase after being exposed to the
autonomous shuttle, and then intention to use shows the highest increases after being
exposed to the autonomous shuttle -- excuse me, to the simulator.
The Whitten Group analysis showed us that trust and safety significantly
increased if they were exposed to the autonomous shuttle last, and intention to use
significantly increased after being exposed to the simulator.
So our next step of things to do is of course to complete the study and to
look at gender and age and time effects as folks are being exposed to this technology. This
is also laying the foundation for us to further look at folks who are disabled, those with
dementia, the disadvantaged, the deskilled, as well as those with degenerative disease such
as degenerative neurological disease. Because we do believe that transportation, like Mr.
Leonard had said, does not just need to be affordable, effective, integrated and ubiquitous,
but for these special populations it also needs to be accessible, acceptable, adaptable, and
available. And so, you know, we hope to further the work in that area.
So I want to acknowledge our funders as well as our stakeholders, and then
each one of the project team members, specifically Justin Mason, our post-Doc, Dr. James
Russel, who did much of the data collection, Virginia CC Opeco, one of our collaborators,
Jason Rogers, and our research assistants, Melissa Varara and Casey Clark. So thank you
so much for your attention.
DR. RANKA: So I think we are slightly ahead of schedule, right, so maybe I
can take a few more minutes? All right.
Good morning, I’m Sanjay Ranka, I’m a Professor in Computer Science. I
do mostly artificial intelligence for transportation and healthcare.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
32
So what I’m going to talk about today is the world is changing, where in the
past most of the data on transportation was collected by the government or third-parties like
Google. But now with autonomous cars, all the car companies are also going to be having
data, and that’s the new player in the market.
And the key goal is to able to create a symbolic relationship, and in my mind
the way to do that is make things more real time. Because at some juncture all the cars that
you are driving around, we have to interact with pedestrians and other vehicles. And the
more real time information which you can provide to them, the more amenable they are
going to be to able to use it. And obviously that’s going to improve both safety and
operations.
What we’re trying to do is we’re trying to build a data store. Again, try to be
able to access that in real time, to be able to achieve a variety of applications, both from
monitoring, from evaluation analysis, be able to do predictive modeling, being able to do
optimization of a variety of issues in signals and corridors. And also support planning and
investment.
Again, the key goal obviously is to both manage traffic condition and
improve safety, but make everything data driven. And also be able to support our vision,
movement of goods, services, and people.
With that challenge I’m going to talk about some of the projects which we
are working on on trying to make the projections smarter and more real-time. The goal is, at
a very high level, to be able to compute everything on the edge, as they call it, right near the
signal, as close to the signal as possible, because you want to make it real time. And then
be able to assimilate key findings and push that through a centralizing monitor like a cloud
so that you can do more longer analysis there. Clearly the important part here is to be able
to be sole sufficient and be able to use a variety of sensors which are available to you today
and which will be available to you in the future. And that’s what we are working towards.
So one of the projects which we are working on is what I call smart
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
33
intersections. So right here basically is that you have an intersection, a typical intersection
in any city, which has local controlled data, there are local controls is under the roads, which
give you data about every vehicle arrival. We want to be able to use other cameras, like
video cameras, which are called fish-eye cameras, sitting on top of the intersection, and in
the long term be able to use LIDARs and other sensors to be able to collect information in
real time about all that’s happening on the intersection, store it in a sensible form, keeping
track, making sure that we obviously worry about privacy issues. And then be able to
understand what’s going on at the intersection in terms of an anonymous behavior or in
terms of near misses, to be able to provide feedback to the designers as well as architects of
intersections on what designs are actually more effective and what signal timing planning
are actually more effective to improve both operations and safety.
So what we’re trying to do is we’re trying to develop these real time skillable
systems which can actually collect data from roof controllers as well as from video. First I’ll
talk a little about the roof control of it and then I’ll talk about the video analysis. To be able to
collect the data in real time, to be able to run all these algorithms for detection and
everything else in real time, and then be able to push that information back to the
autonomous vehicles in real time for this to be effective.
So we are developing a system for understanding what ’s happening at the
intersection, both at the intersection level and at city level. We work both the City of Orlando
and City of Gainesville, collecting permission from all the signals which they have data for, to
be able to analyze it. And I’ll show you some of the results from a broader perspective, and
then go back to the intersection level.
So for example in this particular case we can actually process one vehicle
feed in less than an hour. And that’s our goal. Our goal is to make everything as real time
as possible so that the information which is gathered is captured and processed and
analyzed and sent back to the vehicles and the pedestrians. We can actually take all the
data in a given city, rank all the intersections in terms of their performance so that traffic
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
34
engineers can understand what’s going on in every intersection and get a broader picture of
– I’ll give you an example of this. Where essentially what it’ll do is it’ll rate, if you look at one,
it shows the different categories of intersections, whether the fixtures are not working
properly or they have capacity issues, or they have timing issues. Again, the intent basically
is to provide a traffic intersection with a good perspective of what’s going on in that entirety
so that they can make changes in real time in some cases, or a case on a daily basis.
So they can select all the information about what’s going on in the city for
any given day, be able to figure out the week day or weekend behavior, to say hey, this is
where my problems are during the week days and the weekends, understand what’s
happening in the morning, what’s happening in the evening, so they can plan the traffic
signal timings better.
Again, the goal is to make the infrastructure smarter so that they can impact
whatever the conditions and autonomous vehicles.
This one kind of shows you that the behavior changes by the hour, and how
you plan for that. Also be able to do a change detection automatically to say where the
change is happening so you can focus on those changes.
This kind of gives you an idea in terms of a broader infrastructure. What I’m
going to talk a little bit about is how do we do real time video analysis to be able to
understand what’s happening at the intersection and, again, be able to generate messages
on the fly and be able to support in a few years information through the autonomous vehicles
about what information is useful for them, what’s happening in the city or whatever unit it’s
in, and our intersection can be changed without direct ly.
So this one kind of shows you again real time processing of an intersection
in Gainesville where traffic is coming. We’re tracking different kinds of objects. In this
particular case I’m showing you in real time all the cars, the greens are a bike or a motor
bike, and the yellow is pedestrian. So all this tracking is happening in real time. Also this
kind of gives us a good synopsis of what’s happening at the intersection. We collect this
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
35
data in real time and then process it to understand behavior. You see this is just playing
again and again.
What this data does it give us information about all the vehicle categories on
that particular intersection. So it would be color coded for different directions, the light, all
the categories are marked so that the light ones where they start and the dark ones where
they end. And this shows you the behavior at that particular intersection. Again, the goal is
to be able to analyze this and be able to rate all the intersections in terms of their safety and
operational quality.
This shows you the behavior on the right-hand side based on different
phases, generally every signal has six to eight phases. And each phase has different traffic
behavior. In some cases people are going east/west, some cases people are going
north/south. And then be able to also understand what’s happen on a given phase because
traffic sometimes moves in a fashion that two directions can interact with each other, so we
can actually show that. We actually then are able to analyze also all the pedestrian traffic
through the intersections to see how many people are actually jaywalking on the
intersection, how many people are not following the paths. And again, we can do this from
morning times to evening times.
A very interesting part which we do is we look at autonomous behaviors,
right? So for example these are all the categories here which show the way that you are
crossing on the red light in a given hour. The key goal of all this is that once we can analyze
this data, we are able to then predict behavior and then be able to use that for better
intersection design. And in some cases then the end goal is to, if you look at the left, there’s
a pedestrian walking, crossing the traffic and goes back, right? So we want to be able to
capture that kind of behavior and be able to warn all the traffic on that intersection to say
hey, there’s a pedestrian crossing where they should not be crossing, be aware of it.
So these kind of real time messaging and real time processing is very
important to improve the safety in the long term. And the goal is that any object which can
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
36
actually communicate at the intersection should be able to get that piece of information.
Here’s another example. If you look at the behavior of this particular
pedestrian, they come there with the red light, doesn’t want to wait, comes back, and jay
crosses. So we want to be able to track all this behavior in real time and be able to send this
information to all the vehicles on the intersection.
We also want to be able to detect near misses. So if two vehicles are
coming very close to each other we want to be able to say hey, there’s a potential problem
there. And the key challenge of course there is that you’re dealing with esthetical cameras
which have distortion of all kinds, they see better near and they don’t see very well far away.
We have to be able to take care of that. And then again, be able to find out. So here’s an
example on a target, you see there that the vehicle stopped in the front so the vehicle
behind braked. And if it didn’t brake there would probably be a crash. And the vehicle could
possibly brake in this location because there was actually a fire truck behind which was
beeping, so the vehicle in the front stopped in the middle of the intersection and the one
behind went further.
Another example of that, if you look at these two vehicles, they come very
close to each other but that’s normal behavior. Because they’re just passing each other in
different directions. They came very close to each other but they’re in different directions
and that’s not a near miss, so we have to be able to detect one, things which are real misses
versus “the false misses” if you will..
So we do this analysis by collecting the data in real time. And then part of
our project, and Lily mentioned this, to be able to do this in the competing project where we
will have all these video cameras at 10 to 123 intersections, we’ll collect the data and be
able to essentially share the data with the smart and connected vehicles in real time of
what’s going on in the entire network. Be able to analyze the network itself, detect all the
incidents which are happening on the different parts of the intersection off their network, and
relay that back to autonomous vehicles and be able to optimize it at the network level.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
37
We also do other work on the infrastructure site where we understand why
looking at vehicles which are passing on highways to be able to say what commodity are
they carrying. One of the interesting parts of understanding behavior on highways is to
understand how commodities are getting transferred in the State of Florida. And obviously
this can be extended to the entire nation by essentially looking at the type of object and logo
detection of what company it is. So if it’s Wal-Mart we know what kind of goods they are
carrying, if it’s Fed Ex we know what kind of goods they are carrying. So we can understand
in real time how goods and services move around in terms of freight planning and everything
else.
The other thing which we do is we look at highway videos to understand
how vehicles enter and exit ramps, how does that impact speed of the different vehicles
which are coming in, to be able then to understand what kind of ramp changes need to be
made or what kind of designs are better than other ones, to be able to be effective.
This is my team, there are four faculty involved in this project and eight or
nine Ph.D. students. We effective double up every aspect of these techniques. Our work is
supported by NSF Services Smart View Project, I’d like to acknowledge them. Also this is a
recent study which showed the importance of different levels in terms of improving
transportation, it just came out six months ago. And UF as a whole is focusing on the top
five here, which can actually have a real impact in terms of improving traffic performance.
Thank you.
DR. ELEFTERIADOU: Thank you. Do we have a few minutes for
questions, I believe. Yes, please.
MALE SPEAKER: Thank you for visiting us. We’re an academic institution
so we’ll give academic reactions to what you’ve said. Four quick ones.
For your presentation, what you’ve discussed about improvements in
autonomous vehicles at intersections could be done now. Nothing preventing that from
happening now. So it would be useful to think a little bit about the political economy of why
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
38
is it we aren’t doing these things now and we need autonomous vehicles to get them done?
I’m not saying you answer it, but it might help to sort of motivate what ’s going.
I’ll go quickly through the three other ones. These are largely for motivation.
On your presentation you show the sort of stylized facts about the declines
in fatalities for VMT. Nothing obviously to dispute about that. The issue is why. You know,
we’re really not so sure about that. I mean one can speculate about improvement in
vehicles, but it’s very hard to show that kind of thing. I would suggest that an equally
compelling explanation is the increase in congestion that’s occurred at the same time
because people are going slower, it’s simple physics. Point being, that’s a benevolent way
to reduce traffic fatalities, is congestion. And the point that I would make is that we’re sort of
running out of good ways to save lives. And that’s where autonomous vehicles comes in,
because that’s going to be a little accelerated with decline, and that’s a good thing that’s
going to happen.
On yours, you didn’t have time probably to go into detail, but it will help to
know more about the people in your sample. In other words, what was their accident
experience? Had any of these people been in accidents? Do they know people who’ve
been in an accident? How much VMT they drive every year, they care about safety in the
vehicles they buy. Going into this, this would be an interesting comparison. Are there
groups of people who we think, because of their own experience to vehicles and so on and
so forth, are going to be open to the benefits of autonomous vehicles and safety as
compared with others who aren’t at all similar to those experiences, who obviously would be
very, you know, hesitant to believe that for a variety of reasons. Might be interesting to sort
of point all that out.
DR. CLASSEN: I will comment to that by saying that we do have that data.
And so that will be analyzed as we are done with --
MALE SPEAKER: Yeah, that’ll be important I think in terms of the people
who resist it.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
39
And on yours, the thing I’m thinking is long run. Is the road network going to
change because of autonomous vehicles? Remember, we have built this network based on
non- autonomous vehicles. And I’m seeing all these intersections and I’m thinking, yeah, we
have that because people couldn’t get close to their destination and they gotta walk across
city streets. So maybe traffic patterns will change, maybe we’ll realize, you know, these
kinds of intersections are bad, roundabouts are better, or for that matter we’d even want
these kinds of things. You know, it doesn’t hurt to start treating now the changes in the
network that will come about because of autonomous vehicles, and start speculating.
Obviously you can’t do any of this now, but I think we’re going to learn a lot as we go into
this and we’re going to see both pedestrians and the network changing. So those are
suggestions, our view.
DR. ELEFTERIADOU: Thank you. With respect to your questions, if I got it
right. Yes, we can start doing this now. One of the issues is the chicken and the egg. The
vehicles need to be instrumented, the infrastructure needs to be instrumented in order to
communicate, and I think that is what is unique with what’s happening in Florida is that both
are happening at the same time and that allows us to start getting those benefits.
DR. CRANE: I was just going to say about the congestion. Okay. All I
want to say about the congestion was that’s a very good point that I hadn’t thought of, but
it’s definitely not what I want to see happen. But that’s an excellent point. Yeah.
DR. RANKA: As we know, intersection unfortunately doesn’t change
overnight, right. So all these highways and road networks are built over decades. So I know
that these things are going to stay for a long time. We have to work with whatever we got
right now. And of course as we learn more, as you’re pointing out, they will evolve. But
probably take several decades before that happens.
DR. ELEFTERIADOU: Any more questions? Yes?
MALE SPEAKER: Urban planner. Strikes me, this is fascinating. The big
data revolution is giving us tremendous resources, but where do we draw the line between
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
40
rational human behavior and irrational human behavior? Because theoretically and totally
unacceptably you can have, you know, collect the entire, you know, data on the entire
history of one pedestrian who is apparently trying to cross the road. You know, the mental
history, is he out of an institution, might he have a relapse, does he have something in t he
family recently? Yeah, we have the data about that, yeah, his wife has behaved -- where do
we draw the data? Where is the human responsibility in that? And not ours, in yours as
scientists, but ours as individuals? Thank you.
DR. ELEFTERIADOU: I think one of the points Sanjay was making was the
real time information and warning. And so being able to recognize and predict intent, that’s I
think where that technology can be very useful.
DR. RANKA: But, you know, add to what Lily just said, the doctor can only
tell you what you need, good for your health, if you still want to do stuff which is bad for you,
at some juncture we can only prevent other people from not getting impacted.
MALE SPEAKER: Thank you. I just saw an article on line where the
Attorney General is ordering that software always be created with a back door so that
government can access it. Of course anywhere there’s a back door in technology, hackers
can misuse it. What are the provisions for encryption security in autonomous vehicles? Will
they have no back doors so that no hackers can intrude?
And to follow up on the previous gentleman’s point, we probably all
remember of the Geico commercial where the squirrels are like high fiving each other for
tricking a car running off the road. What happens when 15 year old kids see autonomous
vehicles come along, decide to kick a soccer ball across the street and watch what
happens?
DR. CRANE: The second part, I think that’s already happening, you know.
These are just anecdotal stories that I’ve read on line. But people challenge the
autonomous vehicle, they’ll drive right in front of it because they want to see what it does.
And that’s a problem.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
41
The encryption part I’m probably going to have to pass to Sanjay myself
because the level that I’ve been at is just got to get it to work. And that, you know, becomes
kind of my secondary concern. But, boy, if it is hacked into it could just be chaos. I don’t
know, do you have any comments?
DR. RANKA: Yeah. You know I don’t really know much about the security
of autonomous cars themselves, but on the infrastructure side we are making sure that
whatever we actually collect we throw away as soon as possible. We don’t, like for example,
the video coverage I talked about, we just need it for five seconds, and after that we process
it and at that juncture the video is useless for us. We just wanted to get immediate
information about what is going on. We obviously are very careful about privacy issues, and
we encrypt all the information which we keep. We analyze it, that even if you were able to
get to it, it’s useless for you. Because our goal is not to track people, our goal is to
understand behavior.
FEMALE SPEAKER: Thank you. My question’s about the connected
infrastructure and the infrastructure pieces. Because I know in the autonomous vehicle
community there’s an actual competing theory, including one that’s being tested right now in
Florida at the Villages, where they want to be completely independent. They don’t want th is
connection. They don’t want to have to depend on the infrastructure being there doing what
it’s supposed to. They want the vehicle to be smart enough and the mapping to be good
enough.
So I want to sort of talk to you about if you take on what this debate is about
connection versus true autonomy.
DR. ELEFTERIADOU: Yeah. This is a very interesting just competing
approaches. In my view you need to have both to maximize safety and to avoid making
congestion worse.
I don’t think it can be done without the connectivity, just purely with
autonomy. The information that you can get with connectivity supplements very nicely what
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
42
autonomy is able to do. So in my view the two have to work together.
DR. CRANE: As I say, it all started like 2007 was the Urban Challenge, and
you didn’t want to rely on anybody, it had to just be you. And part of that though is look at
the impact across the infrastructure. So we had no impact on the infrastructure, and yet you
could get it to go.
But the benefits are if there’s other vehicles, especially pedestrians is one of
the projects we’re doing now, is that if I can be told that from sensors at the intersection or
from other vehicles or whatever, that information’s just really going to help me. So again,
when we started, well there is nothing, so we’re just going with the current infrastructure.
But I think anything that’s added is just really going to be a benefit.
MALE SPEAKER: So a good segue way for my question is so if these
autonomous vehicles are relying on the connected infrastructure, and as we all know that
things break at one point in time. Where is a line that you will draw in terms of how
autonomous systems are supposed to be just self-sufficient like because otherwise people
will start taking shortcuts and, you know, build systems relying on the infrastructure that may
fail and may get some potential hazard.
DR. CRANE: I was just going to offer my opinion is I think it ought to be
self-sufficient, and anything I get in from some of the other sources is a bonus for me, it’s
going to make me better. But I think it’s going to have to be, and as it al l develops too, I
mean we’re in northcentral Florida, and you don’t have to go very far to be out to not much --
well, I’ll take that back, to, you know, minimal things. And I think self-sufficiency is going to
have to be with us for a while.
DR. RANKA: So by definition most autonomous systems are failsafe. That
means they will work independently, maybe at a lower quality than expected if the
information is not there, and the goal is that they have become more efficient as more
information is available. But the intent always is that things should work even if everything
else fails.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
43
DR. ELEFTERIADOU: Yeah. I think there has to be redundancy. I also
think that the deployment will be facilitated if both work together to supplement the
information that the autonomous vehicle has right now. For the examples that I showed, for
example where you don’t have line of sight to obstacles, but having information from the
infrastructure allows you to get that.
Okay. Over there.
FEMALE SPEAKER: I’m sorry I was late. Thank you. I was wondering
about a timeframe. A lot of people say this is going to happen much later than what we
think. And maybe this is another group, but supposedly the 5G network is very important.
And China’s way ahead of us, and we’re arguing with Huawei and are we going to use them
or not use them. Supposedly it’ll be a security risk.
And also trucks. Now those are high paying jobs, truck drivers. Are you just
doing vehicles now or will they be doing trucks later? I’ve heard there’ll be one lead in truck
with one person and then autonomous behind it. And I’m wondering if a young person, not I,
but if the young people want to get into this field, what’s their best major in college, or what
should they be studying?
DR. CRANE: I just tell our students that the whole area, in the last part, is a
mix of mechanical, electrical, computer science, and that you can major in any of those, but
you’re going to learn about all of those. So it’s a blend. Maybe it needs to be a new major, I
don’t know.
ER. ELEFTERIADOU: Yeah, it’s traditionally the transportation systems
has been under civil engineering, but you see here the collaboration that happens at the
research level, and we want to bring that also for the students so we have transportation
across different colleges, across different departments, and it becomes so much more
important now with advanced technologies at the forefront.
DR. CRANE: It’s funny, you go to different -- well, an electrical engineer is
going to say it’s all electrical, they’re number one, and the computer person --
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
44
DR. RANKA: I would say that the world is changing where you could take
one profession and work there for 35 years and you’d retire. I told my two daughters, just
learn how to learn. And after that you’ll do fine.
DR. ELEFTERIADOU: All right. We are officially out of time. I put up take
aways from this session from the different presentations. And we welcome more discussion.
Thank you very much.
MR. GOPALSWAMY: Good morning. Howdy.
My name is Swami Gopalswamy. I spent several years in the automobile
industry, primarily the areas of control, design, and development. And for the last few years
I’ve been with Texas A&M University working on connected autonomous vehicles.
So before I go on with what I’m going to talk about, let me introduce my
colleagues and co-panelists. Dr. Darbha is a professor in the Department of Mechanical
Engineering also at Texas A&M with focused on controlled theory and applications,
including, of course, autonomous vehicles. He has been working on vehicle platooning for
over three decades and has developed some fundamental (inaudible) that provide insights
into platooning and connected vehicle operations.
Dr. Rathinam is an associate professor in mechanical engineering also at Texas A&M
University. His research interests include autonomous vehicles, aerial, ground, and
underwater autonomous vehicles, (inaudible) algorithms. Somebody asked in the last
session, what kind of education would make sense to be in this space? Just to give you an
example of Dr. Rathinam, he has his bachelor’s and master’s in mechanical engineering, a
master’s in electrical and computer science, a Ph.D. in civil engineering, and his Ph.D. thesis
work was on aerospace. (Laughter)
Okay. It is a great pleasure for me to be here to introduce the concept that
we’ve been calling -- let me just see how I should -- okay. So, the concept that we are
calling the infrastructure enabled autonomy and the idea is how that leads to autonomy as a
service and what implications could be thereof.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
45
We believe it is a new paradigm for autonomous vehicles and the primary
goal would be to ac celebrate the deployment of autonomy, autonomous transportation. So
let me start with this. Why are we working on autonomous vehicle? Many people have
already addressed that before, and so I don’t want to belabor that point. There are
tremendous benefits in terms of the potential for lives saved, number of crashes that can be
reduced, the economic savings, the fuel savings and so on and so forth. Obviously, great,
great benefits.
But we always look at the fine print. The benefits accrue with the
corresponding penetration of autonomous vehicles in transportation. You need to have 90
percent of penetration to kind of get the fabulous benefits that we are hoping to get. And for
example, in the U.S. alone, we have 250 million automobiles, both commerc ial and personal
cars and trucks on the road today. And the question is, how do we hope to get the kind of
penetration that we need to really get the benefit that we all hoping that we will get? And so
that’s really the thought process behind what I’m going to be talking about today.
We talk about the penetration of autonomous vehicles, and let me kind of do
a quick rundown of what people have been thinking, and I want to do this by using the
famous Gartner Hype Cycle for emerging technologies to kind of quickly walk through what
people have been thinking.
In 2011 -- I don’t know if people are familiar with this hype cycle but
essentially it just talks about the various emerging technologies that are there and the
different phrases of enchantment and hype about this technology before you kind of -- you
go into the (inaudible) and finally kind of get into the plateau of actual meaningful
deployment.
Okay. So in 2011, autonomous vehicles did not have a space here. In
2011, we suddenly started having autonomous vehicles, and as we walk through, you can
see at the peak of its hype around 2015, and whoa, it started going down on the trough.
And, in fact, in 2018, I think -- I don’t necessarily agree with that but the Autonomous Stuff
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
46
technology was split into two parts for different levels of automation and one was continuing
to be in the hype cycle but the other one is going down (inaudible).
So the point is, as I said, we have 250 million vehicles on the road. How do
we expect to accelerate the penetration of autonomous vehicles? Because that’s the only
way we can actually get the benefits that are being touted that we hope to get from
autonomous vehicles.
So I come from the automotive background, and so I have an acute sense
of what it takes, or I have an acute awareness of the realities of complex code because as
you can see, mechatronics, the modern automobile in general is a mechatronics system.
The autonomous vehicle is much more of a mechatronics system and the complexity is all
big into the software that goes into these vehicles.
So we have in a typical modern car, 10 million lines of code. The
complexity is humungous. The number of subsystems, number of engineers working on it,
number of suppliers involved, the distributive nature of the development, so the complexity is
really enormous, and this is just for regular modern cars, not necessarily autonomous
vehicles. And if you just look at some of the headlines, you just have to do a google search.
And this idea is almost a couple of years back, there have been dramatic failures across the
globe in terms of software-related failures, and when you have software failures in
mechatronic systems, they are systems that work not only the software that works with
mechanical systems and that which has interaction with human beings, the cost impacts are
humungous.
So I’m not going to belabor the impact of the complexity except to say that
we really need to think about how soon do we really expect the people to be taking -- for us
to be seeing 90 percent penetration? Even 10 percent penetration, 25 million vehicles on
the road, I expect that we probably will have 20,000 to 30,000 vehicles in the next 10 years
in a very limited kind of fashion.
So here’s my explanation. I want to say how to think about this.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
47
So if you look at the driving functions that we do today, you can look at the -
- let me see if I can get a pointer on this. Yeah.
Okay. So the typical modern vehicle, you have a driver aware system that
is developed by the automotive odium. They help you, once you press the car -- I mean, the
gas pedal, it translates that into how much acceleration you need to get, how much braking
you need to get, how much steering you need to provide. You have the decision-making --
before that you have the situation awareness. The driver is you. You would see what’s
outside. You look at your dashboard, kind of make an assessment of the situation, and do
your decision-making. And then you use, in order to drive, additional information that comes
from outside.
The distribution of the responsibility amongst these functions, the
responsibility really translates to liability to a large extent, is as follows. The automotive
odium takes care of the green box. The human driver takes care of most of the decision-
making and situational awareness. And you have some exclusions from the odium onto the
decision-making by providing some supporting ADAS functions, advanced driver assistance
systems, such as lane change, obstacle avoidance, and so on.
The information comes from third-party players, such as GPS, cloud
services, et cetera, and there are infrastructure providers, infrastructure primarily being
information infrastructure providers who would facilitate the transfer of the information to the
vehicle to do the driving. And this is a balance, a distribution of responsible data that’s
evolved over time and we understand what it is. We care comfortable with it and this is kind
of where we are right now.
If you wanted to move to a new paradigm of completely autonomous
vehicles, what this changes to is the following: The automotive OEM really takes it a
responsibility for the entire driving. And this is where I believe there’s going to be a big
pause in the introduction of autonomous vehicles in a big way. They have already seen
many accidents of that nature if it were to happen for a large company selling millions of
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
48
vehicles, it’s going to be a big, big liability for them.
So I feel like this will provide a lot of big pauses, and so at least, we will
have 20 to 30,000 vehicles. Maybe on a least basis they’re owned by the OEM and
provided under very limited circumstances. And that does not bode well for a goal of 90
percent penetration.
So what is our proposal? Our proposal is to separate this intelligence that’s
going to be needed for autonomy -- for autonomous driving from being only in the vehicle to
being separated between the vehicle and infrastructure. Take advantage of the
infrastructure in a very fundamental way and that would reduce or at least distribute the
liability and that is kind of what we want to propose as a concept as an additional in-panel
development along with the existing approaches for autonomous vehicles.
So the idea is that we have the traditional driving that is taken care of by the
automobile OEMs that will continue to remain there, but the situation awareness will come
from the infrastructure. When I say from the infrastructure, I mean primarily from the
infrastructure. And the decision-making will be done by third party units that reside between
the driver aware system of the vehicle and the situation awareness on the infrastructure to
help make the autonomous driving decisions.
So the core concept is what it considers a separation of issues and
expertise. The automakers will focus on the drive-aware capabilities and make it even more
perfect. They can, of course, bring in more advanced driver assistance systems but then the
infrastructure will become smart and start enabling awareness. And when I say smart
infrastructure, what we mean by that is that the perception and the contextual awareness are
synthesized, will be synthesized using off-board sensing. So you have roadside units that
will have sensors that will generate, that will be looking at -- so you essentially have ice --
not ice on the road but placed on the infrastructure. They will be looking at the road from
above, getting -- generating the situation awareness information. Sending that wirelessly to
the vehicle. And in order to do that we leverage connectivity. That’s the -- there was
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
49
another question sometime back, which one should be used? Should it be connectivity or
should it be autonomous driving? We believe it has to be a combination of both. There
have been significant improvements both in connectivity as well as autonomous capabilities
in the recent past and we really need to bring them together.
And what this allows us to do very interestingly is this allows us to do on
demand automation, and I’ll talk about that later on because that actually suddenly opens up
a possibility for funding this infrastructure investment required that would not be otherwise
possible.
Here is a cartoon that kind of describes our concept vision. So you have the
specialized corridors that we will call as the infrastructure enabled autonomy corridor. This
will have roadside units. They may look like streetlights but in reality they are multi-sensor
smart PATT. Multiple sensors that can essentially look at your road and generate situational
awareness information. And, of course, you don’t want to collect all information and send it
blindly to the vehicles, but you have -- it’s computing. Computing right there on the edge to
abstract that information to the minimally required mode of information that can be sent
optimally to the vehicle in a wireless fashion. Then that information then comes into the
vehicles which have the driver aware capability, the smart connect device inside, which can
talk on the one hand to the infrastructure. On the other hand, to the vehicle’s drive
(inaudible) system, and use the information to generate autonomous driving capabilities.
Once you have that information, it does not have to be used only for
autonomous driving. You could imagine a plethora of uses for such information. It could be
sent to a bicyclist or a pedestrian on their smartphones or a smart watch to give them
warning signals of what’s happening on the roads. You can send it to manually -driven
vehicles such as diagnostics and information so that they can drive better as humans.
So this is the fundamental concept that we have been trying to promote as
infrastructure enabled autonomy. And let me first walk through a few slides that explain how
we have bene taking this concept at Texas A&M University.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
50
So we started by actually having a clarification on the concept by doing a
theoretical explanation of the risks using a probabilistic characterizations on the falls and
risks associated with distributing the intelligence. So we demonstrated that the risks get
distributed according to the -- based on the proposed redistribution of intelligence where we
assume that there’s a clear partitioning of the functional and physical interfaces. Dr.
Rathinam and I actually co-authored a paper on this, and the paper reference is right here.
So that’s kind of laying the theoretical underpinning for why we would want
to work on this concept.
The next one was our iteration zero work. This was where we said, okay,
let’s now do a virtual simulation of this concept to see if we can demonstrate that things
could work. So, we had a simulation of an autonomous vehicle with a three -- and some of
our students are working on it right here actually in this room. We have three cameras,
three roadside units that are simulated and the vehicle goes through each of these cameras
and the vehicle is able to detect -- I mean, the camera system is able to detect the vehicle,
localize that, send it wirelessly. And for wireless communication we use DSRC units. And
the reason we use DSRC’s unit is not because we have a preference for the DSRC or 5G,
we are actually communication agnostic, but we were donated 30 DSRC units for free by
(inaudible) operation. And that’s the reason we use that.
But the idea was that the fact that we could use the DSRC implies that that’s
actually one of the lower bandwidth communication capabilities and we have been able to
successfully use that to demonstrate our system and the implication being that if 5G had to
come on board and if you adopt 5G it’s just going to be that much easier to do it. So that’s
the iteration zero.
The iteration one was development of a prototype at our RELLIS campus.
Is this the recent presentation that I sent you? Sorry. Okay. This is -- okay,
I just lost the title of the presentation. I don’t see it anywhere here.
Okay. So this is their first iteration. And essentially what we have is we
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
51
have a RELLIS campus. This is all a military Air Force base that we are using. It’s got
seven-plus miles of concrete runways that we use as our innovation proving grounds at
Texas A&M. It’s called the RELLIS campus. And that’s where we did our initial prototyping
experiments. You can see this mobile trailer that has polies sticking on top of it. We have a
camera mounted on top of it. We have wires coming down. We have a laptop serving as
the Edge computer and this is the autonomous vehicle that we use to evaluate this
capability. So that was the first iteration zero.
Then the iteration two -- before that I want to show a couple of more videos
because -- let me -- yeah. So this is how it looks like. For example, this is the trailer. The
camera is somewhere here. There is an Edge computer here that communicates to the
vehicle wirelessly through DSRC. And you can see from the camera we can actually track
the vehicle and localize the car in terms of -- in the runway and send it to this vehicle.
Yeah. This is what I want to show.
This is Texas A&M research. So we have to have the hat. But this is our
very first experiment where we were just doing simple lane changes and so on.
This is the RELLIS campus that I just mentioned. It’s a 2,000 acre old --
sorry, not old -- 2,000 acre military airbase repurposed for advanced technology research.
We have seven-plus miles of connected concrete pavement. We have traffic intersections,
specialized roadside units, and 5G connectivity in the plan. And this essentially becomes
our smart city environment that we want to work with.
So if you look at the IEA development, how it has moved up to now, on the
iteration two we have built a 450 meter course with four roadside units. These are
essentially luminaires, streetlights on which we have mounted the cameras. And we -- let
me see if I can -- yeah. And this is connected through a wire to the Edge computer at the
bottom. We are still using the laptops for now because we are still in the development
phase. But this can be now made to switch between DSRC or 5G or any other
communication method. And we have used this for a variety of testing. And that’s kind of --
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
52
some of the testing is what you will be seeing later in the presentations also. And so we
have four of these light units about 50 meters between each of them. And these are not
powered so we actually had to have generators for each of those even though we are doing
our experiments. But this kind of is the beautiful landscape that we get to work with.
So some recent tests. So this is multiple views. You can see the first
camera tracking the vehicle and then there’s a handoff that happens at some point when the
vehicle comes into the view of the next camera and then there’s an overlap period. And
after the overlap period is over, the second camera, second roadside unit takes over and
they are constantly processing the information, localizing the vehicles, sending that
information back to the autonomous vehicle.
The autonomous vehicle has a multiple set of sensors because it is being
used for a variety of research purposes. But for this particular project they didn’t use any of
the onboard sensors. So they used the light exclusively on the information coming from the
infrastructure because that’s kind of what we wanted to demonstrate.
So what I want to say is this -- the concept obviously needs really early
stages, but we have already been able to show enough progress that we feel compelled that
it’s now really time to take it to the next stage.
So the technical value proposition is that the multiple sensor smart PATT,
you can get much higher, reliable, and precise information because you know the ground
route on these roadside units. So you know the ground route. You can have markers on
your road. You can have reconciliation between different roadside units. Within a roadside
unit you can have multiple sensors that are fused with each other. And not only that, you
can also have the ability to award occlusions. So if you are driving today, if your view is
blocked by a vehicle -- a truck or whatever other vehicle that might be in front of you, you
really cannot do much. And the same thing applies to sensors that mount on the vehicles.
You can only do so much with the sensors that you mount on the vehicles. And this is
exactly where sensors, the different vantage point dramatically changes the clarity of your
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
53
situational awareness.
And then the fact that you have increased computational time, you can do
much more in terms of understanding what is happening to your traffic. You can spot
potholes. You can identify traffic jams. And most important ly, every vehicle will have the
same sort of belief states. Shared belief state that is common between all the vehicles so
they know how to try to respond to each other. And mind you, I am talking about a scenario
where it is a mixture of manually driven vehicles and the IEA-driven vehicles.
Now, this is an interesting part. What this whole concept allows is it
provides a new paradigm for doing autonomy. The deployment model. How exactly are we
going to do autonomous vehicles?
So my expectation is the following. I will be driving a vehicle that looks like
a normal vehicle. I would be driving it in general as I usually do except when I go to these
congested areas that have an IEA corridor, my commute time for example that I spend one
hour every day, that’s when I say, no, I will actually want to go to this IEA corridor. The
corridor will interrogate me as soon as I enter there. Say do you want to be driven
autonomously? Of course, there has to be a registration process prior to that. (Inaudible)
that’s done. So it interrogates, it says, do you want to be driven autonomously today? And
if I feel like it I say yes, I want to go here. And it drives you autonomously. And at that time,
because the vehicle is in complete control, you can actually go to sleep. And if you are
asleep when it comes close to the destination, it will park you and you wake up and drive
when you are ready to take it.
And you get charged by the mile. And that’s the key point. So because
autonomy does not have -- and I do not want to be driven autonomously every time. I want
to be driven autonomously only under some conditions. I value autonomy only under some
circumstances and the value of autonomy for me is different, and therefore, ideally I would
like to pay only for the autonomy when I really need it. I don’t want to pay for all my sensors
and costs up front. Here is an opportunity where I pay for my autonomy as a service. I pay
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
54
by the number of miles that I travel autonomously.
So this I think provides an opportunity for monetarization of autonomy. And
this only allows private infrastructure players to come into this space because there’s going
to be tremendous investment cost associated with instrumenting roadside units every 50 to
100 meters with sensors and having connectivity. Yes, of course. But there is also an
opportunity to monetize the usage of that. And that’s kind of really the opportunity that we
want to talk about.
So it’s not only a technical value proposition. So there’s a value proposition
for a large ecosystem. We believe the automotive odium will be happy because they don’t
have to deal with taking on additional liability while they still remain in this space of
autonomous vehicles. Private infrastructure operators, tollbooth operators, cellphone
operators, cell tower companies, lighting and urban infrastructure companies, a whole s lew
of companies suddenly get an opportunity to participate in this autonomous ecosystem.
Device makers, the people who make the sensors. The people who make the integration of
the sensors that go on to the roadside units, opportunities for them. And the application
makers. So when you have a smart connect that gathers a whole bunch of information, the
number of applications that you can have is limited only by your imagination. I believe that
there will be a whole slew of people who will be making different types of applications,
leveraging the information that comes in while you’re driving.
Now, of course, law enforcement, infrastructure/maintenance, traffic
management, a variety of users can easily be seen. And of course, society at large because
we believe that this would actually accelerate the penetration of autonomous vehicles at
least in specific corridors. So on a corridor by corridor business, you can dramatically
improve the penetration of autonomous vehicles, thereby reaping the “promised benefits of
economy.”
So that’s kind of where I want to stop. I mean, there are many open
research questions. And we don’t have anything solved. I mean, we have ideas, and we
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
55
know the problems exist. We are working on them. But really, there is a lot of work still that
remains to be done. We have demonstrated the core functionality at Texas A&M but we
want to increase the complexity of the scenarios that we will be demonstrating. We want to
increase the scope of the work we are doing. We want to go from 40 meters to a two mile
loop. People are working on resilience with respect to cyberattacks to communication
failure. There are a variety of opportunities for doing research on pricing models, perhaps
elastic option-based price models. Dr. Caroke is an expert on that. Then there is a
Goldman Private Investor Partnership models that need to be looked at. There are legal risk
and regulatory environment implications that need to be discussed. A variety of business
models. I’m not going to go through all this.
And last but not least, the issue of privacy. How do we ensure that in
having thousands of cameras all around the road does not infringe on people’s policy?
So I say with that I leave and I’m going to pass it on to Dr. Rathinam.
This is just -- yeah.
(Applause)
MR. RATHINAM: I have been a faculty member at Texas A&M for 10 years.
So the first time, the first year at least when I was, you know, I would come to the lecture
room. I would start talking. I would go to the board, write equations and all that. It was an
undergraduate class so after, you know, some of the classes, the undergraduate students
came to me and said, Professor, please. You know, you can say whatever you want. Just
say howdy before you start anything. (Laughter) And we are all good. So I’m going to say
howdy and start my presentation.
I’m going to talk about infrastructure-based advanced driver systems,
economic systems. So we have already heard a lot of presentations about introduction to
autonomous presentations and the IEA concept.
I just want to first acknowledge the support by the Safety Transportation
Research Center, which is funded by the U.S. Department of Transportation. And also, we
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
56
work closely from the researchers from the Texas A&M transportation industry. And, of
course, all these students who are here, Abhishek, Vamsi, and Kenny, who do all the work.
We are just present; right?
So I’ll just move on to -- so I think we have heard from the previous
presentations at least the road infrastructure is going to significantly benefit both advanced
driver assist systems, which is what is called as ADAS on autonomous vehicles. And in
particular I’m going to talk about proof of concept demonstrations of the IEA concept that Dr.
Swami was talking about for different applications first. And then I’m going to talk about the
need for common standards. So this is something I think Ken, earlier in the keynote speech
mentioned that we need common standards for evaluating autonomous vehicle technologies
and taking it forward. So that is something I’m going to talk about. I’m going to maybe focus
on lane departure warning systems because that’s closely related with infrastructure.
So before I move on to the actual slides, I want to give you a short of intro
video on all the testing that we do. So we work on a wide variety of problems from
platooning, car platooning, truck platooning, lateral control, vision-based control, many other
things. So this is just a summary of what you have done. So here, this is a promo video and
all the frames that you see, the car frames that you see were all taken when the car was was
in the autonomous model.
(WHEREUPON, a video was played.)
MR. RATHINAM: And I am not really a fan of this music but they said it’s a
wide audience so I just said, okay. Why not?
Anyway, so I’m going to talk about this in two parts. The first part I’m going
to talk about the roadside camera unit which we have installed on each of the four poles that
Dr. Swami was talking about. And how we use that information for some of the applications.
And then we’re going to talk about lane markings, which is an integral part of the lane
departure warning systems that at least are available in some of the high-end cars.
So first, in fact, just to give you an idea, deep learning algorithms have been
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
57
used, or at least have been significantly developed over the last few years, and you can use
them to identify objects in the road and you can actually use them to also identify -- classify
objects into cars, emergency vehicles, policemen, and all kinds -- pedestrians and all kinds
of things. And you can actually sort of tailor the rest points based on what kind of vehicles
you identify.
So this is just a short clip of having some of these deep learning algorithms
trained on the datasets that we have. And you can actually classified them into emergency
vehicles and all that. Obviously if it’s a human, manly driven car you would see an
emergency vehicle and try to yield to that vehicle and stop or take the appropriate action.
You can also try to emulate that using an autonomous vehicle. And in fact, that’s what we
have done here.
So this is essentially a roadside unit. Here you have two cars in this
demonstration. You have the French car, which is an autonomous car and you have
another car that is following, which is an emergency vehicle. So the autonomous car either
does not have a sensor to actually know that there is an emergency vehicle behind or it’s
just like anything else that you do. You know, you have sensors but it’s just broken. Right?
It just doesn’t work.
So in this case at least what we have is the information from the emergency
vehicle descends from the camera unit and then that information is sent back to the
autonomous car using the DSRC communication system in real-time and you do the
appropriate response. So that’s the -- what we do here that we are going to show. And this
is all real-time processing. I can go into more details if anyone is interested, but this is really
for two vehicles. We also try to do a platoon of vehicles. In fact, we have two cars -- we
have a platoon of three cars here. In this demo, actually, we have the front brakeless
autonomous car. The vehicle in the middle is a human driven car. And the last vehicle,
which is basically either it’s an emergency vehicle or a police van or whatever it is. And here
you see that there is no data decline between the autonomous car and the emergency
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
58
vehicle. And so these are the kind of applications where the infrastructure of roadside units,
the cameras could really help.
So again, the concept is the same, so the information -- this information is
identified by the camera units and then sent back to the autonomous car which is what is
processed in real-time. And that is the response that you see here where essentially the
autonomous car comes to the other lane, and of course, a manually driven car also parks
itself and you let the emergency vehicle go. Right? So these are some of the kinds of
things you could do.
We have also done that with different test case and here we have, you
know, modeling humans is always the most challenging thing to do and you obviously don’t
want to test these kinds of things using real humans so we have these human-like robots
that you can manually control. We can get them onto the lane and off the lane and all that.
So the concept again is the same. The information is basically sent from the camera back to
the car using the DSRC unit and you can actually post a lot of problems. And I know Dr.
Swami is working on different applications on this.
So this is just an introduction to what is possible when infrastructure is not
involved. I think the point again I want to emphasize is that I don’t think there is one versus
another. I think it has to be collaborative. You need maybe some onboard sensing because
many of the cars that you already buy already has its own GPS units. It has immunosensors
so you can use that information and you can mix it with what you get from the roadside units
and do better things and hopefully improve safety and everything.
And that’s the point here. Right? So I can complement the onboard
perception units. The infrastructure can actually provide real-time information that when you
don’t have line of sight, direct line of sight and it provides redundancy. And essentially, you
know, most of these applications, I’m sure that anyone who works on it knows that things fail
and that’s the standard. The rule of thumb is that things are going to fail. And how do you,
you know, accommodate and react to what is important? So if you have more sensors and
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
59
more redundancy all the better.
So anyway, so that’s the first part of my talk.
The second part is going to concentrate on lane markings, which is really --
are an integral part of the lane departure warning systems. You know, we have taken huge
strides in autonomous vehicles in the past decade and this is going to continue and we are
going to see deployments of these autonomous vehicles in some sections or some cities in
the United States and things are going to improve probably in some sections of the
highways. But we are not, I mean, autonomous vehicles, at least in my lifetime, I don’t see
that replacing -- completely replacing all the human-driven vehicles.
So at least over the next 40 years, you know, you’re going to have human-
driven vehicles also and they have to coexist with autonomous vehicles if at all we have any
penetration. So which means that at least my assumption is that as long as there are
human-driven vehicles on the road, you are going to have lane markings to guide us. Right?
So if you look at it, in 2000, there was less than one percent of all the vehicles at least with
advanced driver assist systems. Depending on what study you look at, by 2030, 50 percent
of the vehicles are going to be equipped with advanced drive assist systems. And when I
say these advanced drive assistant systems, they could be adaptive cruise control,
automatic braking, lane departure warning systems. It could be tire pressure monitoring
systems, things like that.
The lane departure warning system is one of the key features that is
proposed. And the main reason why it is proposed is because people think, at least many
studies have shown that it can actually improve safety.
Now, this is just a report from one of the projects funded by USDOT where
they show that the lane departure warning systems can actually save -- that can potentially
reduce hundreds of thousands of crashes every year and can save a lot of lives. And
there’s around 3,600. And anyway, I don’t want to go through all those numbers but the
point here is that, yes, there is potential for safety benefits but these systems are not yet
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
60
used. And there are a combination of reasons why they are not yet used. It could be
because of lack of good lane markings. It could be because we don’t have good algorithms
yet. Or it could be a lack of common performance standards for evaluating these algorithms.
So if you look at a lane departure warning system, it’s actually just, you
know, I just want to tell you how complex this problem is because, you know, this is just one
feature that I would say it’s a new term technology that’s going to be there in the cars. And
maybe now it’s in the high-end cars but in the next 15-20 years I’m sure it’s going to be
there, more prevalent in most of the cars. And you can see how difficult the problem is when
you also have to rely on infrastructure.
Now, if you look at a lane departure warning system or a lane detection
systems, it really consists of three parts. You have the software, you have the hardware,
and you have the infrastructure. The software is really the algorithms. Hardware, if you use
a camera. So cameras, the reason we don’t talk so much about cameras is because it’s
cheap. You know, the lighters, I think someone talked about it $60,000. I mean, you can
pray and hope that things -- costs come down but it’s going to take time to actually have cars
that you put radars on, really good radars. Right? Cameras are cheap and that’s the reason
why they are an integral part of these lane detection systems.
And of course, you have the lane markings, which is an integral part of the
infrastructure. But you can actually get hundreds and thousands of variants based on what
kind of algorithm you pick, what camera you pick, and what kind of lane markings on the
infrastructure you have.
For example, if you just take off lane markings, you know, they could be
made of different types of materials. You can have different striping processes. You could
have white followed by black and you could have white and black juxtaposed. You know,
you could have different payment types. So you know, it doesn’t matter how good your lane
marking is. You know, if you don’t have contrast between your lane markings and the
surrounding parts of the road, your departure warning system is not going to find it. So it’s a
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
61
contrast that is very important.
And also, your lane geometry. Your thickness and everything. You also
have different cameras you can pick, and of course, the algorithms, many people have been
working on that and I don’t have time. I think I’m only done. I have a couple of minutes.
But the point here is that we are not at the end of the story because the
performance does not just dip in on the algorithm you pick or the sensor you pick or what
lane marking you have. It also depends on the environmental conditions. You know, the
edge cases, which is going to really determine it’s not the usual normative cases. The
normative cases, you know, you can get 90 percent, 95 percent accuracy, which may not be
good but it’s still better. But it’s the age cases that are really going to determine how good
these vehicles are going to be or how good the systems are going to be.
Now, you know, you can have, you know, direction, streetlights that could
radiate your performance. You could have -- you know, this is one of the hard cases. You
could have sun and reflection from the red surface. Snow you could have all markings,
construction sites. Payments.
So really, I don’t think the camera is going to solve all these problems.
You’re going to have multiple sensors and do sensor fusion and really key point here is that
we do not know what the optimum performance we could get out of just these cheap
cameras at this point in time at least. There’s a lack of common standard which could help
us in keeping track of the different components -- the evolution of the different components --
the evolution of the different components of the system and how it affects the performance.
So that’s what really we are working on. We are trying to develop at least an open source
platform where we can test the infrastructure side, the algorithms, as well as the hardware
and all that. And we are also developing metrics to actually evaluate and rank these
systems and do it. Of course, it has benefits to the federal agencies. The USDOTs and the
state DOTs because they can actually answer quest ions like do my lane markings in my
state actually -- are they fit for a level two operation or a level three autonomous vehicle
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
62
operation. So you can answer questions in a simple but basic questions like that. You
know, obviously, it’s useful for infrastructure providers because they can actually test new
materials and algorithm developers, of course, because you get new sets of data you can
work with.
Anyway, I’ll just stop with this. I just already running out of time. But the
point here is that you need a systems approach. It’s not about just developing algorithms or
it’s not about, you know, the infrastructure companies just doing their lane markings. It’s an
overall system so you have to look at everything and see how the improvements in each of
the parts will affect the entire system.
So with that I’ll just stop. Thank you.
(Applause)
MR. DARBHA: Hi. I’m going to depart from the standard howdy.
So anyway, so I have about 10 minutes so I’ll try to make it crisp and short
and I’ll also answer some other questions that were raised earlier, each in connection with
what all I’ve been doing.
So I’ll focus on certain aspects of connectivity and infrastructure. So for me,
infrastructure comes through communications.
So what I have been doing for 30 years depends primarily, it’s mostly based
on how information amongst vehicles will help promote safety and mobility. So this is
basically what the focus has been on, and I’ll talk a little bit about emergency lane change
then. But I’ll take this time to tell you something about one of the questions that was raised
about autonomy and connectivity.
So my interest in this particular area of autonomous vehicles goes back to
my doctoral thesis, which is about developing automatic vehicle following control algorithms
in the ‘90s for the for the Automobile (inaudible) Systems Project. So I’m also interested in
what all -- not just in developing the technology but also assisting what the impact is on
traffic safety and traffic mobility.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
63
So to give an answer to that question, so if you are looking at adaptive
cruise control systems, which I think all of you know, so the minimum time gap that you need
to maintain, which again goes back to my thesis, is that it should be at least twice the sum of
all the reaction dealings, the action lags and dealings that you have in the system.
Otherwise, you would have this slinky-type instability, accordion type effect, which will cause
pileups.
So recently, so this roughly translates for trucks it’s about 0.6 to 0.7
seconds. So you’re looking at a minimum time gap of about 1.4 seconds. But for (inaudible)
and other kinds of things, people are interested in a sweet spot of about 0.7 seconds if I’m
not mistaken. So to cut down that much with the app which controls systems, it’s just not
possible. And people have been doing experiments, especially in the T&O group, as it was
in the California PATT group, they assured that if you had the cooperative adaptive cruise
control system, that is like if you had the information of the vehicle and had acylation
information of the vehicle, you could cut down the time headway.
So, in fact, my recent paper, this is a paper that I’m quoting, it basically
assures that it can carry down by a fact of two if I had the acylation information of the vehicle
ahead. But if I had the information for all vehicles, I had like let’s say three vehicles ahead, I
can cut down this minimum time gap by a factor of four.
So that means that connectivity does help and here is where the
infrastructure can actually play a significant role. For example, this is under the assumption
that you have perfect communication. Okay? So vehicle to vehicle communication will not
help in the imperfect case because it’s primarily local because autonomy is biased on
onboard sensors. But if you have connectivity, you can get it to the infrastructure what is
happening much farther ahead. And that is what is required, and that is the way how we can
actually make the benefits quantifiable.
So, and the same thing with safety also. There are a lot of safety studies
that one can talk about.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
64
I also work on -- the infrastructure also plays a significant role in commercial
vehicle operations. For example, one of the projects that I’m doing, the Safety Center,
Universal Transportation Center, is on developing automatic diagnostic systems for
autonomous trucks. Essentially, the purpose is to facilitate pre-trip inspections, automate
them. You could also use it for automating enforcement inspections and to make sure that
trucks comply with the FMVSS 121 standard. So this is the kind of thing that we are doing.
There is some work that infrastructure can also help in traffic harmonization,
some work that I’ve done recently, and then in this particular work for lack of time I’ll just
basically focus on emergency lane change.
So how does infrastructure help in emergency lane change maneuvers? I
think as the previous speakers of both my colleagues here and the Florida speakers have
mentioned that you could get real-time information about vehicles, obstacles in a time-
efficient manner. But more importantly, it’s not just that that you require for safety. It also
requires some way to localize your positions very effectively. And here is where
infrastructure can actually play a significant role.
So normally in terms of -- I also want to mention here there are two types of
queue that you would have for lane change maneuvers. Here in the first one you would
have the queue for the emergency lane change maneuver comes from the front of the string
of (inaudible) you have. And in the second case, as my colleague has mentioned, if it’s like
an ambulance or something it actually comes from the back of the string.
So if you had information about these things much earlier, then you could
plan and act accordingly. And in the second case it’s even more important because it could
save lives. So that’s the reason why it’s fairly important.
So in the emergency lane change maneuver, essentially, it’s a maneuver in
which you have an obstacle in the road and the vehicles are basically trying to evade this.
And here I’ve shown a two-lane scenario where the obstacle is in the right lane and the
vehicles are basically trying to move over to the left lane. Okay? The point of this entire
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
65
exercise is to show that if you did not have the information of the obstacle and there’s no
communication between vehicles, then the following distance that each vehicle should
maintain should be much more than otherwise. The reason is, the vehicle behind would be
blinded by the vehicle ahead because it will be blocking the obstacle and the thing is you
would want the following vehicle to see the obstacle at least at the same location that the
previous vehicle, the preceding vehicle would have seen it. Okay? And this cannot be
possible unless you had some sort of communication or you keep a huge following distance.
And that’s basically the point of it.
So the rough calculations will tell you that it should be, at the controverted
end, it should be about 1.8 seconds of time they need to maintain. And that’s clearly not
acceptable from the mobility point of view.
So I think I don’t want to talk a whole lot about the experiment and facility at
RELLIS because I think you’ve seen this before. So here what we are doing is we are
looking at the emergency lane change maneuver design and control of design. So the
difference between regular lane change maneuvers and emergency lane change maneuvers
is that the vehicle when it makes a lane change maneuver, you cannot make use of the lane
markings of any of those things because the trajectory of the vehicle ahead is designed in
real time. And what the following vehicle has is only a limited preview information about how
much of it it is. So the question is whether you can actually make the design happen and
have a controller which actually tracks the path of the vehicle ahead. This is essentially
what the problem is.
So we simulated, because I’m only dealing with one vehicle from the point
of view of safety first, and then after that, probably in the next year or two we’ll do the two
vehicle and three vehicle lane change demonstration in a similar setting.
So in the first run, what you would have is we just had these cones, and
then we had the vehicle drive manually to collect, restore the GPS information. And this
GPS information will basically define the trajectory for the second run, which is the
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
66
autonomous mode if it’s the following vehicle, and what we want to do is we want to track
this. And the amount of information that is available to the following vehicle is just a few
samples of where it is. That’s about all that you have.
And you can actually simulate it for multiple vehicles, too, because you can
store the second vehicle run and then use the first two vehicles and then you can see what
happens to the third vehicle.
So these are technical details. I’ll just skip it.
And this is based on the work of my student, which we presented in the
ECC in Naples.
So I have only one second left. So I’ll just show some demonstrations. So
this is at 45 miles per hour. I’ll just show two vehicles. This is the aerial view at 45 miles per
hour, and the other one would be at 70 miles per hour.
So the cones are the ones which are defining what the lane -- that is the
double lane change maneuver.
This is an in-car view at 45 miles per hour.
None of the cones fell over.
Okay. This is at 45 miles per hour. So I’d like to show you the 70 miles per
hour videos and then I’ll stop after that and I’ll take any questions.
So the drone that we have was not able to keep track of it, so that’s why you
see it blocked because this is going too fast.
And this is the last video. This is the video. This is at 70 miles per hour.
So I think I’ll just pass at the moment. So thank you and I’ll be glad to take
any questions.
(Applause)
SPEAKER: Thank you for the interesting presentations. I like the way you
emphasized both the opt-in sort of paradigm as well as thinking in systems. The question I
had was in my Chrysler Pacifica I have all the latest advanced safety features you describe
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
67
with adaptive cruise control. And I’ve got Android Auto that’s pulling in the data feed from
Google with speed limits, congestion, other things. But currently, those two systems don’t
talk to each other. So I’d like to just take the speed limit feed from Google and tell my
adaptive cruise control, just go the speed limit that you know is there. Do you see that kind
of like system connection emerging in the coming years or are there different incentives for
the OEMs and the data providers that is kind of hindering that or what do you see as how
that might play out? Because then you could have adaptive speed limits if the weather is
bad and that would feed into the adaptive cruise control. How do you see that evolving in
the coming years? Thank you.
MR. DARBHA: This is a problem that is considered -- this is traffic
harmonization. So this is a traffic harmonization problem. That’s where people usually refer
to where it’s like variable speed things in different sections of the highway where you advise,
that’s the kind of the thing that is considered a traffic harmonization problem. I think this can
be done. It’s not -- the question is, how do you set the speed limits on the highway
depending on the weather conditions? And there is research that is going on and there is
some preliminary work that I had done last year with some of my colleagues. So it can be
done.
MR. GOPALSWAMY: I just wanted to add one thing. One is
technologically whether it can be done, and that’s of course like what Dr. Darbha is saying.
It’s definitely feasible. But the bigger question is will the OEM trust the data? And if the data
is going to come from a completely different source, then the question of trustworthiness of
the data becomes an issue. Whereas, if there’s a prior agreement with the infrastructure
information provider that this information is trustworthy in some sense, then I think it’ll
happen. Until then, you cannot, you know, do it and make people delay on that.
MR. WINSTON: So normally in scholarly work we begin with a literature
review. Now, in this area this is quite difficult because you’ve got a lot of players. There are
many universities in this country and all over the world that are working on autonomous
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
68
vehicle problems. There are many technology companies and auto companies working on
this. And even government are working on all these problems. And I think it really would
help this presentation if you stepped back and gave us an overview, really, that this is a
cooperative and competitive effort. Because what you’ve got, its’ in the weeds. You’re one
player. Maybe you’ll be around in 10 years. Maybe you won’t. And I’ve seen these
presentations, and what I’ve seen is much more sophisticated than what you’ve got. But
then they had a lot more time to present it.
And my point is, it’s useful to put this stuff in context. In other words, what is
distinctive about what you’re doing as compared with the big picture which everybody is
doing? As I said, this is a global, cooperative, and competitive effort. But at the same time,
those features are what are so important to actually giving us optimism that we’re going to
get these things sooner rather than later. You may not be part of I but Stanford might be,
Ohio State might be, Michigan might be, or no universities at all. That is , it’s all driven by the
private sector.
So I think broader perspectives on really what’s involved here, and not just
one small thing and one small problem. You mentioned environmental problems. There are
tons of companies who have identified and are working on all these problems as we speak.
You know, it’s not that you can continue to point to things . I mean, everyone is out there
doing this and I think it’s important to really get that message across.
MR. GOPALSWAMY: I think that’s a fair point but let me explain two points.
Two parts to it.
Number one, most of the work that’s being done is either looking at
autonomy by itself, focusing on the ability for autonomous vehicles. Or using the
infrastructure to provide augmented information, such as in traffic intersections and specific
places like that. There has not been work that we know of that is focusing on providing real-
time information from the infrastructure that will actually be used to drive the vehicle
autonomously without using sensors onboard the vehicles. So that’s actually -- we have
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
69
done the research and we have not seen any work that does that. Number one.
Number two, on the part of the environment, the people recognize that there
is an environmental issue. People have been working on the front types of snow weather
conditions. But what’s not happened is taking a systems view and saying we need to have a
standardization that will provide a metric that can be used to say, is this system good? Is
this infrastructure that we provide, is the lane marking going to provide adequate support for
lane (inaudible)? Is the camera system going to provide adequate support for lane
(inaudible)? That’s the systems-level standardization that is (inaudible). People have been
focused on what is important for them. For example, the people who do vision, they are
focused on vision algorithms. The people with the cameras, they are focused on the
cameras. But there is not a general purpose here for it and that’s understandable. That’s
kind of what we want to say and trying to say we would like to have lane keep, lane marking,
lane detection standardization, a metric system that can be used to promote what it means.
MR. RATHINAM: I’ll just add one more point. At least from my point of
view, most of the work that has been done in companies is from the point of view of single
vehicle autonomy. That is, you have -- try to have all the sensors on board and try to make
your vehicle as safe as possible. I think one of the efforts that we are trying -- I mean, we
didn’t have time to present today but one of the things that we are doing, Texas A&M is
trying to look at safety problems and other issues for the collection of autonomous vehicles
when they are connected. So if you have a platoon of vehicles that are driving on the road
and they’re all talking with each other, of course, if you have communication, great
communication between the vehicles it’s all good and fine, but what if that doesn’t happen?
What if there are communication failures? You know, let’s say if you have a collection of five
vehicles, if there is a communication failure between the first vehicle and the third vehicle
and fourth, whatever it is, how does it affect safety and performance? And so I think our
effort has been focused on that.
Companies try to do that. I don’t want to name the companies here. They
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
70
have tried to develop safety metrics from a single vehicle’s perspective, but it’s all great but it
would not work in reality. The reason is that if you’re just looking at single vehicle and just
trying to make that safe, I don’t think you will get the efficiency benefits that you desire. So if
you want safety and if you want to have the kind of performance that people are talking
about, then you need to look at the collection of vehicles and that’s something that we are
looking at and I don’t think the companies are -- I mean, those are long-term problems,
research questions, and companies, they may get into it but not right now.
MR. GEDDES: Yeah, I don’t want to jump the queue, but thanks.
So Rick Geddes from Cornell Public Policy.
And I want to go to a point Professor Gopalswamy -- do I have it right? --
has touched on, which is super important in the policy aspect of infrastructure. And that, as
you mentioned, charging by the mile. One of the biggest challenges seems simply, is how
you pay for the infrastructure. And currently we pay on basically a per gallon fossil fuel tax
on either gasoline and diesel fuel at both the state and federal level to pay for, you know, the
vast majority of the roads. But you mentioned charging by the mile. The State of Oregon is
the leader in pilot projects over the past decade to try to make that simple switch from
charging per gallon to charging per mile. And it’s been a real slog, a real challenge
politically to get drivers to accept that, although the State of California is interested in doing
that.
So there’s my questions for anyone on the panel. Do you see the advent of
autonomous vehicles as facilitating the acceptance -- and they call them mileage-based user
fees, MBUFs, in a way from a per gallon gas tax so that people will be more accepting of
incorporating what we would think of in terms of utility pricing as a facility charge not the
price of a trip. Right? So it’s been sort of driver resistance. That seemingly simple change
which is super important in the economics of this, could we see, you know, the interaction of
autonomous vehicles and the way you pay for the infrastructure as facilitating that
improvement?
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
71
MR. GOPALSWAMY: Yeah, my thinking is yes. And the reason I say that
is if I look at toll roads, that’s the prime example where the, for example, in Texas, we have
all these -- and in fact, in a lot of places we have this auto charging of tolls and it’s become
so ubiquitous that you start forgetting and then you start not worrying about it too much as
long as it seems reasonable and it is actually in some ways based on usage.
So my point is people would be more comfortable paying by mile if the
process of doing it is very easy and very convenient. If it’s seamless for them, the almost
don’t have to worry about it. Then I think it would be okay against if you actually have to pay
explicitly for a bill.
SPEAKER: Thank you. First of all, I wanted to build off of Cliff Winston’s
question. And I think first of all, I think it’s not true that there are no companies that are
working on independent infrastructure dependent autonomy. I am aware of two companies
that are building their sensors into the infrastructure because their use cases, they want to
serve a limited service area of multiple areas so it turns out to be more effective to build the
sensors into the infrastructure than to build the sensors into each vehicle. And so I think,
and so I’m aware of two companies who are building out that approach in commercial
service. And so I think that gets to a larger question which is, what is the right mechanism
for public-private interaction and particularly private academic interaction? And I’d like to
turn this into a question, and the question is, what has your mode of interaction with the
private sector been over the last five years in which we’ve seen not just one but two C
changes in the direction of investment in autonomous vehicles. About five years ago, we
started to see a massive amount of private sector investment in autonomous vehicles.
Brookings put out a report that said about $80 billion had been spent on JVs and M&A in the
autonomous vehicle space. And then so in the last year or so, we’re seeing increasing
consolidation and engagement in the academic sector. So I would hope that the approach
around how do you divide problems, which ones are best approached by the public sector
and which are best approached by an academic sector and which ones are best approached
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
72
by the private sector has changed. So I’m curious about what you’ve seen over the last five
years or so.
MR. GOPALSWAMY: It’s a pretty complicated question. So let me try to
answer that.
My thinking is that first of all, some of the work that’s done by private
companies, we don’t have access to that. All we know is they are a marketing blurb. So
from a literature survey perspective, we will not be able to assess what exactly they are
doing. And so my answer was more in the context of the universities. So that’s number
one.
Number two, but on the flip side, there have been companies, you are right.
For example, I think Daimler has a parking lot scenario where you actually drive inside a
parking where there are a whole bunch of sensors in infrastructures that support that.
What we have been working on is the extreme scenario where we are
saying, okay, how much can you push it? Can you be completely independent from sensors
to be on the vehicle and use almost 90 percent of the information coming from the
infrastructure? That is kind of how we have been trying to position ourselves.
In terms of the public-private partnership, it is actually an ongoing thing. For
example, we are discussing with some companies, specific companies which do lighting, for
example, who want to get into smart infrastructures, smart cities. They have been willing to
support it. In fact, we may even actually have a formal agreement with them to support
some of the work.
We have had conversations with many other companies such as 5G
companies that want to install 5G because they need a physical infrastructure also and we
have the RELLIS campus, and so they have been willing to come down and say, okay, we
will install 5G. Can you demonstrate use of that in a very meaningful application? So that’s
something that we are working on.
We have talked with automotive odium. They have expressed interest but
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
73
we have not had actual work yet.
SPEAKER: I just wanted to make an on-the-record comment for basic
research and investment in basic research. All of you sitting there, 30 years ago the
investment at Berkeley and other places in intelligent control system, PTT planning,
computational geometry, the whole PATT program at Berkeley. So I think we have to very
much compare (inaudible) the fact that investment in basic ideas, risky ideas , 30-40 years
ago that took place, is not paying off and you people are talking about just certain aspects of
it, but basic investment was made 30-40 years ago and you should try to keep as a policy
matter in perspective that fact because in the present environment you will see that we are
only picking up the fruits but we are not looking at the roots of the thing. I think we have to
pay attention to that fact.
MR. DARBHA: Yeah, again, I’m one of the early graduates of the PATT
program. So that was not my piece the program was on. So thanks for finding that out.
People seem to have forgotten the PATT program.
MR. RATHINAM: I also work for PATT. And you know, we -- we work on
fundamental problems and that’s where my interests are. Today’s presentation is just one --
I had 15 minutes to talk so you know, if you give me more time I can talk over the
fundamental problems. I’m not sure how much of the odium I can retain but I’d be happy to
talk about it.
SPEAKER: I don’t know how much time you have.
MR. GOPALSWAMY: We don’t. Sorry.
SPEAKER: Thank you.
(Applause)
MR. RAY: Okay, howdy. Great. Welcome back from lunch. I hope everyone had a
nice break. I’m going to introduce our keynote speaker for the day.
Derek Kan is here. He is the executive associate director at the Office of
Management and Budget. He has a long history of working in the transportation industry
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
74
and in policy. He was most recently undersecretary of transportation for policy at the
Department of Transportation, which he held almost for two years. He’s also been a
member of the Board of Directors at Amtrak. He was a general manager at Lyft. And then
he had several other private sector experiences before that.
Derek also worked in the Bush White House where I knew him. He got his
MBA at Stanford GSB where he was an Arjay Miller scholar. I can tell you I was co-teaching
a class when I was getting my Ph.D. at Stanford more than two decades ago, and the Arjay
Miller scholars are the top 10 percent of the class. And I can tell you that I wish everybody
was in that top 10 percent. Of course, mathematically that’s not possible, but they were the
best students. You get that award for the highest GPA in the class, and it was really a job a
teach them.
He also has a degree in economics from LSE and he graduated from USC.
So, without further ado, here is Derek Kan to talk about autonomous
vehicles and the topics of this conference. Thank you. (Applause)
MR. KAN: Thanks, Korok, and thank you all for being here. Nothing like a
post-lunch fascinating conversation. (Laughter) So hopefully, the blood is flowing from your
brain to your stomach and we’ll see who stays awake in 45 minutes.
I wanted to thank you for the kind introduction. It’s great to be here at
Brookings. There’s a lot of familiar faces in the room. You had a great presentation this
morning from Ken Leonard, who leads a lot of the work at DOT. And the afternoon is much
more exciting. I wish I could stay because the next panel is fascinating with some of the
foremost experts around P3.
So you have Rick Geddes, who literally is the nation’s top expert in public -
private partnerships. Marjory Blumenthal has perhaps the best most comprehensive
analysis she wrote at RAND on a safety framework. And Cliff Winston has this really
creative, fascinating paper on why traffic deaths drop during recessions. And it’s really, in
short, a self-selection process is sort of what he predicts. But just a great set of panelists.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
75
And at the end of the day you have Tanya Snyder chairing a panel. And
Tanya’s great, she’s a reporter at Politico. She always asks very insightful and provocative
questions. I’m thankful she’s not going to be able to ask me any, I hope.
So I want to do a few things this afternoon. I wanted to share some broad
observations on mobility and broad trends. Then I wanted to talk about broad observations
about autonomous vehicles, what the Federal Government is doing for mobility and AVs.
And end with just a lightning round of big questions that folks are asking and my short
answers.
In terms of broad observations on mobility and autonomous vehicles, we
see a megatrend around mass urbanization. The data shows that about 55 percent of the
world currently lives in major metropolitan cities. By 2050, that’s expected to grow to 66
percent. Consequently, urban congestion will be the greatest challenge for the next
generation.
And you not only see mass urbanization growing, but also the concept of
mobility and the concept of a car is changing. One of the megatrends around the car is they
expect cars of the future to be three things: shared, connected, and electrified. And I t hink
the jury’s still out on some of these, but shared, certainly the case; connected, a lot of the
panelists this mornings talked about the importance of connected vehicles and it seems like
there are demonstrated safety benefits; and electric, I think that one has some open
discussion and sort of open debate if we see that.
We’re also seeing with mobility a much greater understanding and insight
into travel patterns like we’ve never seen before. If you look at how people travel, most
recent data from the Federal Highway shows that people -- travel patterns follow a Poisson
distribution. Seventy percent of the trips you take are 10 miles or less. And once upon a
time, you had three modes of transportation: mass transit, car, or train. Well, then the
airplane came along and that sort of was the three or four main travel modalities. However,
today you have eBikes, you have scooters. And what you’re seeing is for these 70 percent
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
76
of trips, there’s going to be a whole set of options that you’ve never seen before.
And on top of that, with companies like Uber, Lyft, and DiDi and all of these
ride-share companies, we now have an app that can help us make the trade-off between
time, distance, and price, which is something we’ve never had before. And as a result, the
way we travel is changing rapidly.
Also, people tend to think that transportation will move slowly and evolve
slowly. That’s not completely true if you look at history. In 1900, there were about 8,000
cars in the United States. Twenty years later, in 1920, there were 8 million cars in the
United States, a thousand-fold increase. And from 8 million cars in 1920, today you have
about 270 million cars. And so the first 20 years of a new technology you see radical
transformation. And today, we have more cars than we do adults over the age of 18.
Part of this, part of this fascinating expansion in the early 1900s, was the
creation of the car dealership. And I’ll touch on that briefly. But from 1900 to 1920, tens of
thousands of car dealerships grew across the country. And as we started thinking about the
autonomous vehicle, there’s a core policy question of this important constituency, car
dealerships, what will be the future of them?
We’re also seeing new business models that didn’t exist a year ago, from
ride-sharing scooters to eBikes. And as we think about new business models, dealerships
are going to play an important role.
Two other sort of macro trends here. There’s significant regional variation
as you look at automobiles generally. Detroit, when it was first conceived, was the perfect
mix of talent and supply chain. It was where a lot of railroads cross and so it was easy to get
auto parts into that region. And suddenly, you had automakers, parts suppliers,
transportation hubs, what we in Silicon Valley call the ecosystem. You had a perfect
ecosystem to distribute this innovative technology at the turn of the 20th century.
However, in the last few years, we’ve started to see a massive shift away
from Detroit to the South in states like Alabama, Kentucky, and Tennessee. And so today,
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
77
Alabama is one of the largest manufacturers of autos because of things like tax incentives, a
robust ecosystem of talent and suppliers, and close proximity to a supply chain, namely
ports and transportation hubs.
And so this raises an important question, I think, for policymakers, for
researchers. Where will the first 10,000 autonomous vehicles be produced in this country?
Or maybe around the world, in what country will you see the first 10,000 vehicles? And in
what state will you see these first vehicles? Will it be in Silicon Valley, where technologists,
computer scientists, and sensor experts live? Or will it be in a manufacturing hub where low
cost of labor, auto parts manufacturers persist?
Finally, one broad observation, global competitiveness. Right the United
States has perhaps the greatest capacity for innovation and I think what we’re going to see
in the next 10 years is which country will lead around autonomous vehicles. Right now there
are really three broad models in terms of national regulation.
You have the Chinese model, which is a very unitary, central plan model,
where they have the next 15 years’ layout of what milestones they hope to hit in terms of
cars produced, tech milestones. They literally have a national product roadmap for the
introduction of the electric vehicle and the autonomous vehicle.
You have Europe that has a council. In Germany, they have a council, and I
think it’s called the Golden Council, that together this council made up of stakeholders of
unions, of car manufacturers, together set policy.
And then you have the United States, which is very open. We have federal
voluntary policy, some state policy. But for the most part, a very open, let the thousand-fire
blooms approach and see what emerges.
The fascinating thing is going to be over the next decade which of these
three models will emerge. And I have a point of view that I’ll share at the end, but it’s an
important question that we as policymakers think what does the competitive landscape of
this look like?
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
78
Some broad observations on autonomous vehicles. Every 5, 10 years
people have a look forward of what AVs will look like. And I think people love predicting the
future of AVs and have we hit the milestones. In short, we’ve gotten some things right,
some things wrong, and a lot of things we don’t know.
What did we get right? We knew that the technology would be very difficult.
We knew that partnerships between OEMs and technology would be important. And we had
some interesting theoretical observations very early on in the process. For instance, the
introduction of automation and AI taught us that autonomous vehicles actually do something
very different with the concept of risk. And Korok has a great paper called “Driverless Cars”
on the concept of risk.
Autonomous vehicles didn’t diffuse risk, it actually centralized risk. They
centralized risk with the engineers. And so instead of those of us in this room potentially
getting in a car accident, what we did is we resulted in a single computer engineer potentially
putting into a line of code a major incident that everybody in the room would suffer if their
care was in a similar situation. And so that was an early observation that theorists had and
it’s helpful in some ways and something we need to be mindful of was we move forward.
I think some other things we got right, two more, one is partnerships are
key. The Department of Transportation recently put out autonomous vehicle demonstration
grants. And if you look at these grants they are designed to have people apply as partners.
And we’re realizing that no single entity, no city, no OEM, no technology company can do
the entire stack. Rather you need a partnership that manages hardware, software,
customer. All of these things together is going to enable the autonomous vehicle to come to
market.
And finally, we also got right I think that the city, state, and federal rules will
be very different. And I think for the most part these are largely agreed upon. They’re not
settled matters, but they’re largely agreed upon.
The Federal Government’s role is to oversee the engineering and the
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
79
design of the vehicle. The state’s responsibility is around the operation of the vehicle, so
speed limits, things like what is considered safe and unsafe behavior, and insurance which
has largely been delegated to states throughout our nation’s history.
We got a lot of things wrong, though, too, at the same time. We got wrong
things like timing. We assumed that products would come to market much faster. Every five
years a forecast says, in four years this will happen. We’ve actually it seen with all the
OEMs. And literally every prediction of when cars will be on the road have been missed.
Ten years ago we thought there would be thousands of autonomous
vehicles deployed in the country today. That certainly hasn’t been the case. Today there
are 1,400 vehicles all being tested in the United States, but none that are being sold for
commercial purposes.
Number two what we got wrong were metrics to evaluate success. And I
hope this afternoon the people in this room can really help us with this question. It’s a very
complicated policy question. But if you recall, when autonomous vehicles first hit the road
there were two main metrics people loved to talk about. Maybe three, but two main ones.
Number one, vehicle miles tested. Whose tests did the most miles? Who’s
hit a billion miles? Who’s hit 2 billion miles? Who’s hit the most miles tested? Number two
was disengagements. How often did a safety driver have to take over the car?
Disengagements, miles tested were the two metrics that some states asked autonomous
vehicles to report on.
Now, we’ve learned over the last few years that both these metrics, while
helpful to advance the dialogue, create major principal agent problems. Vehicle miles tested
is a very, very blunt test. Think about it. If you drive 50,000 miles across the coasts, in
between, that’s a very different test scenario than 50,000 miles inside a major metropolitan
area. In some parts of the country you could drive a thousand miles without seeing another
person, another car. And so this whole concept of vehicle miles tested, we sort of look at it
like, yeah, that probably wasn’t the best metric to lead off on.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
80
The second also created major challenges and took us a little bit for a
curveball. That was disengagement. All cars today have -- or a vast majority of cars today
have a safety driver. So if you’re driving in a vehicle and there’s a problem, the safety driver
takes over, takes control of the vehicle. Some states, city of San Francisco, state of
California, I believe, wanted people to report on the number of disengagements.
Well, the problem with disengagements is it’s a gameable metric. And this
is one of the points that Marjory puts out in the RAND paper that you’ll probably hear about
in the next panel. But if you can game a metric, if you can game a safety metric, it’s not a
very good metric. And so as policymakers, it’s something we spend a lot of time thinking
about, what are reasonable metrics, non-gameable metrics?
And we actually see it. It’s not simply an autonomous vehicle problem. It’s
a problem that exists throughout all transportation. For instance, dangerous driving. We
know the consequence of dangerous driving has a clear red line, which is a fatal car
accident. But precedent to that, what are other examples of dangerous driving we have
proxies for? But if it doesn’t lead to a fatal accident, was that really dangerous or not?
And so you get into these challenges of what are good proxies for safety
knowing the outer bounds of safety are accidents and the outer outer bound are fatal car
accidents. So one of the things we got wrong very early on were metrics to evaluate
success. And it’s not anything wrong, I mean, in the evolution of policy. It’s where you’d
expect us to be.
I think one other thing that we are moving away from are the SAE five levels
framework. We talk about Level 3, 4, and 5. And we’ve realized that Level 3 is an important
-- it’s a useful framework for the starting point, but Level 3 is a very broad bucket with a lot of
examples. And we’re starting to figure out how do we parse Level 3 SAE testing.
What we still don’t know. I think there are two big things we still don’t know.
One, what are the metrics we use to evaluate success? And this is so important. I truly
hope that the people in this room can help us think about this question and think through
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
81
what should the metrics be that the Federal Government uses, that a state government uses
to measure success.
And number two, what is the sensor stat? That’s a technology problem, but
a very serious one. There really are two big heuristics that exist in this day and age. One is
the autonomous vehicle replicates two core human functions, core human cognitive
functions: the ability to see and the ability to think. And the theory goes if the ability to see
is what the autonomous vehicle has tried to replicate, then all we need are cars. We don’t
have internal radar. We don’t have internal sonar. We just have two eyes with which we
can see. And so the concept is let’s have only cameras, not LIDAR and not radar and not all
these other things, simply a camera.
The other end of the spectrum is all of these things are important. We
actually do have depth perception, which is radar. We do have the ability to see how fast
things are moving, which is LIDAR. And so to replicate the eye you can’ t simply use
cameras. You need a comprehensive sensor stack of LIDAR, radar, cameras.
This is an important decision because it really gets to who is going to win in
the hardware space and starts getting to regulation and policy. To what extent should
policymakers say this is what your technology stack needs to look like?
We still don’t know what the stack looks like. We don’t know how to
measure that sensor stack. And we really don’t know yet how safe is safe enough.
Our hope is that where we are as a nation, though, is at a good place, and I
think it is. We currently have 1,400 vehicles in the United States being tested by I think over
70-some different companies. And I think Ken Leonard in the back has all of the numbers
chapter and verse. In California alone, we have 1,200 vehicles being tested.
I want to spend the last few minutes talking about policy broadly, what the
administration’s doing, and conclude with a lightning round of major open questions. The
administration has laid down broad principles on autonomous vehicles and it touches and
builds upon the earlier part of my talk.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
82
Number one, prioritize safety. No matter how exciting this is, no matter how
many lives we can save, no matter how many people in this room can sleep in the backseat
of an autonomous vehicle like they are right now, you know, it could satisfy, safety must
come first.
Number two, tech neutral. The Federal Government nor any government in
our opinion should be saying that this is the ideal technology stack. So we call that tech
neutral. We don’t know if one form of LIDAR, a solid-state LIDAR, is better than a non-solid-
state LIDAR. We don’t know if a LIDAR is better than a radar. Rather what we want to
move toward is a performance-based policy, a performance-based regulation versus
something on the front end where we lay out specific, detailed approaches in terms of the
inputs. So prioritize safety, tech neutral.
We also are not top-down command and control, but rather our approach
has been to modernize regulations. If you look at regulations, particularly the Federal Motor
Vehicle Safety Standards, there’s 400 different -- 4 different chapters. The second chapter
of FMVSS is called Crash Worthiness Standards. The 200s are general Crash Worthiness
Standards. In the case of an automobile accident, what are the internal components and
design standards of a car that reduces the impact and ensures the safety of the people
inside the car?
Well, as policymakers an interesting policy question came before us when I
was at DOT. What if this car is designed to never have a person? What if it would never,
ever transport a human being, not as a driver, not as a passenger? Do we need to put in
airbags, windshield wipers, mirrors?
Well, there’s one company in this country designing a self-driving package
delivery vehicle that says FMVSS, Federal Motor Vehicle Safety Standards, requires these
things, so we have a seatbelt in it. We have an airbag in our prototype. (Laughter) And I sit
and think like where is your airbag placed? Like who are you protecting? And so this is a
perfect example of how we need as a Federal Government to modernize our regulations.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
83
At DOT we’ve looked at all the activities around autonomous vehicles in
three big buckets: regulations, research, and public awareness. On the regulations front,
we’ve streamlined the exemption process , how you can apply to produce more autonomous
vehicles for sale. We’ve begun to update and modernize standards around Federal Motor
Vehicle Safety Standards. And what that means is systemically going through every rule we
have to better understand is it applicable for an era in the concept of things like package
delivery?
And some are easy. Some are very easy and upfront. For instance, does
the steering wheel need to be designed a certain way in terms of the radius of the steering
wheel if you don’t have anybody behind it? And so that’s one clear example where we
probably don’t need things like a steering wheel and brake pads.
But then there are other parts of the Federal Motor Vehicle Safety
Standards that raise other questions, such as in case of crash avoidance, how proscriptive
should the Federal Government be in laying down specific engineering rules as to how self-
driving cars can avoid accidents?
So the streamlining, the exemption process is easy. Modernizing standards
in FMVSS is a little hard. Some are easy, some are hard.
Two other fronts we’re spending a lot of time on is on research. We’ve put
out -- this administration has put out two autonomous vehicle voluntary guidance. Secretary
Chao, when she came into office, was very eager to make sure that the Federal Government
would not be an impediment to the adoption of autonomous vehicles, but rather that would
be left to the public sector, to the public. And that the only real constraint of adoption should
be a willingness of the people in this room to get in back of it. We do not want the Federal
Government to be the constraint here.
Also on research we’ve done things like incorporated new technologies into
our discretionary grants. And we’ve even begun to do things like develop crash dummies for
rear-facing seats and researching alternative occupant protection systems. So that’s
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
84
regulations, that’s research.
The final one we spent a lot of time doing is on public awareness.
Campaigns like this, conferences like this are great because the policy questions are
complicated. And so we’ve done a number of listening sessions and a number of events
around the country highlighting good research, innovation, and also the policy questions that
exist.
And so at a high level what DOT has done in the two and a half years under
Secretary Chao and President Trump is transformed itself as an organization. For the better
part of 50 years we were focused on existing transportation modes. However, we’ve moved
from basic existing transportat ion sources to new technologies, and it’s been a cultural
change like you’ve never seen. As we look at everything from discretionary grants to
activities around the country, we’ll ask ourselves how do we imbue and embed technology
into this? And it’s a long road, but the department is well positioned to embrace this.
We also see this shift and this embrace of technology on the Hill. Last year
Congress appropriated $100 million for autonomous vehicle demonstration grants. And right
now there’s talk of should we enact an AV START bill which deals with federal preemption in
allocation of liability in this space?
I want to end in our last few minutes with a few big questions and sort of a
lightning round and conclude. I see six big questions that folks like asking about.
Number one, where does regulation go from here? I expect regulation to
have a lot of stops and starts. I expect the Federal Government to put out more RFIs,
maybe even a Proposed Rule, get a lot of comments back, and will stop and analyze it for
all. I expect states to do the same. I expect some states to go out very quick, something
happens, and they pull back. We see that in Phoenix right now.
Number two, when will AVs launch? The question everybody always has.
Hard to say, but my guess is much longer than people think. A lot of people are predicting in
the next few years. My personal sense is it’s going to be a few years, five years or longer,
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
85
before we see autonomous vehicles commercially viable for the broad public.
Number three, what will the role of infrastructure be? And I know that’s a
key part of today’s conversation. My sense is infrastructure is one of many variables in
determining what region wins, what country wins. And what I mean by that is infrastructure
V2V, V2I technology is potentially important, but not all car companies believe it’s a
prerequisite to success. Rather it’s going to be one of many variables, such as state and
local laws, such as the ecosystem I talked about earlier, customer acceptance and public
perception, and how you will have fixed and zoned routes. So infrastructure will be
important, but I don’t think it will be the driving exogenous factor.
Number four, how will liability be sorted out? And I think it’s an important
question. We’re starting to see the development of insurance products for ride-sharing
companies. I expect the same to happen with autonomous vehicles. And so how will
liability be sorted out? I think through the market and through the creation of new products
and then finally through the courts.
There’s no clear federal rule for insurance as its generally regulated at the
state level. So I see states starting to think about how they assign liability. And right now
the model people frequently point to is the aviation model.
How will locals respond and what will it look like around the country? I think
there will be a patchwork for a long period of time with local governments learning from each
other. And you’re starting to see that now with ride-sharing and right now you’re seeing it
with scooters. Some cities are very, very much embracing scooters. Others are setting
quotas on scooters. Others are setting licenses. And so we see cities having a very
different approach and then evolving into cities teaching each other what they’ve learned.
And we’re seeing it now with airports. For anybody that travels to an airport
in a ride-sharing car, you’re now taxed, often at the airport, and that airport revenue is used
to make up for parking loss revenue. And this was not sort of a centrally planned thing. Just
a few airports had this creative insight in elastic demand of going to an airport in a ride-
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
86
share. Two dollars isn’t going to make or break it. And so you’re starting to see cities
across -- municipal airports across the country embrace this.
I expect the same with autonomous vehicles, that there’s also a major,
massive business opportunity here. Helping cities develop the platform and tools to manage
new mobility options is going to be something major and something significant that you will
see one day there will be the Google platform of all mobility options in cities.
And finally, who will win globally? My sense is Asia will likely win in the near
term. There the next three, four, five years we’re going to see China just start producing
more and more autonomous vehicles. But I believe in the long term the United States will
win just because there are parts of the technology stack that are probably incorrect that
we’re not even thinking about today. Just like the cellphone. When the cellphone first came
out, which was about 11-1/2 years ago, the camera being a part of it was strange. Why do
we need a camera in our pocket all the time? But now as we go out, I’m going to be taking a
selfie with all of you, so clearly, it’s just such a critical element of our life.
I want to just conclude by saying this is an exciting time and this is just a
time in which there is so much hope and excitement around autonomous vehicles and there
are so many eyeballs on this new technology. And the excitement and the hype in AVs is
shared by the administration, by the department.
But underneath all of this exc itement there’s a huge body of work being
done. There’s coding done by computer scientists. There’s the muscle of automotive
engineers putting components into cars. There’s hardware engineers thinking of what the
next graphics processing unit should look like, the grit of infrastructure builders, and the
insight of policymakers. And I think half this room is made up of policymakers, and this is
where a lot of the hard work will be done.
Believe it or not, the papers you put together, the paper that RAND put
together was such an important piece of work. The driverless cars piece that Korok wrote in
December, so helpful in shaping the views of policymakers.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
87
So for those of you in this room that are sitting here listening and thinking
about how AVs play out, I welcome and encourage you to keep being part of the dialogue
because it’s through the notice and comment, through the Request for Information that you
inform policymakers about what AV 4.0 should look like, what the next set of AV rules should
look like. And so that help is so important to all that we do.
I spent a lot of time today predicting and sort of sharing a forecast of what
the future’s like. In the words of Yogi Berra, prediction is hard, especially about the future.
(Laughter) So take it for what it’s worth.
And with that, I want to thank you all for being here, for being part of this
great conversation. And I’ll open up for Q&A. (Applause)
SPEAKER: Thank you for your comments. It’s very helpful. I know you
had said earlier that you felt that infrastructure is not going to be the driver for AVs. But I
think it’s safe to assume that most AVs are going to be EVs. And so with that, given that
there is money in the public space, particularly through states, to spend money on EV
infrastructure, and I know that there has been some hesitation on spending some of that
money because states are very concerned about I want to spend this money smart, I don’t
want to spend it for something that’s good for three years. I want to think about the next 5,
10, or 20 years with respect to how to spend this money. And on top of that we know that
Senator Carper has proposed a highway corridor bill which could introduce additional
monies, particularly along the highway corridor to support this.
So with that in mind, a highway corridor bill, the VW settlement, a lot of this
money is in state hands, what should they be thinking about today to spend that money
wisely and be able to enable technology that we want to see in the near future?
MR. KAN: The question was how should cities and states think about all of
this money that we’re seeing in infrastructure from the Surface Transportation
Reauthorization bill. Earlier this week, Senator Carper and Senator Barrasso, chairman and
ranking member of the Environmental Public Works, released a bill. The bill was about $297
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
88
billion over 5 years in increased infrastructure spending.
I don’t think infrastructure’s going to be an enabler, but I do think
infrastructure has the potential to be a blocker. If you look online you’ll find a lot of pictures
of situations where self-driving cars are unable to drive in because the infrastructure just
doesn’t make sense. So while the right infrastructure will not bring AVs to your cities or
states, the wrong infrastructure will deter you.
So I do think that you’re going to see from the Department of Transportation
and from Congress an expansion of how surface transportation programs operate. And just
in general, the way surface transportation programs operate is they are federally funded,
state administered. And so the Federal Government takes the money from the gas tax,
gives it to states, and there is a broad set of eligibilities that state can use this money on.
States are seeing increased eligibility for things like technology and electric
vehicles. That was passed in the last Surface Transportation Reauthorization bill called the
FAST Act. Also, there’s a number of discretionary grants that the Federal Government has
and we’ve incorporated technology into it.
So the way I would think about it as cities and states is this Department of
Transportation, this administration is very eager and open to new technologies and
innovation. And so if you look at the discretionary grants that the DOT put out, technology
was an important element.
So what I would encourage cities and states to do is to think how can you
infuse AVs, new technologies, innovation, EVs into the thinking of applying for state and
federal funds? Because it’s definition something that the Federal Government looks at and
it’s something that has increased eligibility.
What will be interesting over the next six months is not only did we expand
eligibility, but are you going to see a shift in criteria where there’s actually emphasis placed
upon EV infrastructure or AVs? My general sense is part of our tech neutrality is while
industry seems to be pursuing electric vehicles and autonomous vehicles, this administration
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
89
does not take that as a given. And so if there’s going to be an AV with a combustion engine,
we’re not going to say that doesn’t make sense. Rather we are tech neutral, part icularly as it
pertains to EVs and combustion engines.
SPEAKER: Thank you very much. My question is what are the real costs --
vehicle’s cost for everyday life, for everyday citizens? I’m not sure that anybody mentioned
about that subject. Particularly I’m concerned like automated streets, you have to pay
(inaudible) how much. Probably it won’t happen too soon, but if it happened everywhere I
wonder that maybe somebody has to think about something like the food stamp, something
like a road stamp for low-income people.
MR. KAN: The question was how do we think about AVs to specific
populations, like low-income and populations that have greater needs? Well, I can share
with you the Department of Transportation was given $100 million. And it’s actua lly
presented a very complicated policy question. In a space where there’s billions of dollars,
the Federal Government gave us $100 million to allocate, well, there literally is billions of
dollars from venture capitalists to OEMs. So the way we look at this pot of money is where
is there a lack of investment? Where is there a market failure in the sense of not enough
research going to a particular area?
The specific area in which we invested it were in populations with the elderly
and the disabled. And we thought that most AV companies are not going to go out and think
how do we serve low-income populations or populations with people with disabilities?
Rather we as the Federal Government, as the administration said this is probably going to
be an area where you don’t see a ton of fruitful research. And so we designed a lot of our
research agenda around this particular question. How will AVs serve this population?
And on this front, and this is sort of going back to my experience before
government, the amazing thing about ride-sharing companies and autonomous vehicle
technology is it is a great tool for certain populations, such as the elderly, such as those that
are in need as the questioner asked. Because we could deliver transportation to your phone
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
90
and tell you please go outside right now to catch a self-driving car that’s going to come and
find you.
And so what you see is this technology has so much potential to really
transform the lives of people. It’s just are we going to have enough investment and focus on
it? We as a government, as an administration are trying to answer this exact question and
make sure there is enough research done to what everything from the human design of the
car should look like to how will we ensure that populations, all populations will have access
to these vehicles?
SPEAKER: So one of the things you brought up was the importance of
states and localities. And following on the prior question, if you are blind or unable to drive,
you literally can’t sit in the driver’s seat, even in a self-driving car. The Uniform Laws
Commission did just complete after two years a model four states to overcome these types
of legacy problems. To what extent do you see the administration’s ability to help normalize
state laws across the country with respect to enabling access and mobility?
MR. KAN: Great question. What is the Federal Government doing to help
mobilize state patchwork of laws around mobility, particularly for specific populations? We
try to find -- this administration’s philosophy is we want to find that right balance from
identifying good practices without being heavy-handed and tell all states to do something,
particularly when it’s not entirely clear to us that there is a federal role as set in the
Constitution.
And so if you look at AV 3.0, the guidance Secretary Chao put out last
September, September of 2018, we actually lay out model or best practices at the state
level. And in that section we talk about specific populations. And I expect at some point in
time over the next two years we’ll update that and say here are what we’re seeing among
states. And here are the best practices, some of the best legislation we’re seeing in states,
and we recommend that other states start thinking about these things.
And so what we’ve done is we’ve tried to pull up at the federal level all the
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
91
things states are doing and put together resources that we can then distribute to other state
legislators. Our view is not to be command and control, in large part because populations in
every city and state are different, so it’s somewhat presumptuous for us to believe that
there’s a one-size-fits-all approach, particularly when it comes to particular populations.
Because that’s not just population, but it’s also what does the city look like? How is the city
designed? And these are things we clearly recognize as delegated -- authorities delegated
to the states.
However, we believe there’s an important federal role in identifying the best
practices, the model legislation, the best practices legislation, and then circulating that and
sharing that with state DOT heads and state governors.
SPEAKER: So there’s no legislative framework for autonomous vehicles.
There are no federal regulatory requirements. It’s all voluntary . And Tesla is saying that
they’re going to have -- release kind of a software update within this year to have fully self-
driving cars.
Already they have autopilot on the highways. They’re saying that they’re
going to have the capability to do the same in urban areas. That’s making a lot of people
very anxious. But is there -- what is the Federal Government’s ability or interest in setting
some guardrails around that?
MR. KAN: Great question. The question was there are no federal laws or
regulations around this. What is the Federal Government’s role going to be when
companies claim or pronounce they’re going to put cars on the road?
So that first part of your question is actually somewhat of a misconception.
The Federal Government does have tremendous authorities. We have something called
defect authority and we have the ability to direct automakers to cease and desist. So defect
authority is a very powerful authority that NTSA has. If there is a component of a car that is
defective, we can pull cars of the road. And our interpretation of our defect authority is that
this applies to the algorithm, the AI, the brains of an autonomous vehicle.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
92
And we describe this in AV 3.0 and our voluntary guidance that we have all
of the authority that we need. That if we see bad actors, if we see cars not doing what
they’re programmed or what they’re marketed to do, we will take those cars off the road.
And, in fact, we have. In Florida, there was one AV shuttle provider that had claimed they
were testing this AV shuttle. Well, it came to the department’s attention that they were
transporting children to and from school or around the city. And so in our mind that is clearly
not testing. That’s actually deployment of technology. And so we ordered them to stop all
testing.
And so the concept that we aren’t doing anything is a bit of a misnomer, but
we actually are monitoring vigorously what is happening across the country. And NTSA’s
defect authority is an extremely powerful authority where we can take any set of cars off the
road.
The hard part of this as policymakers is when there are cases in which it’s a
combination of poor user error and poor design, and it’s not entirely clear where the problem
is. And that’s oftentimes where the gray area becomes should we more aggressively use
defect authority or not? And so Tesla’s going to be one example where we are going to be
tracking the deployment of Teslas very closely and seeing how that plays out.
But there are federal laws. There’s a federal law that has been on the
books for decades that empowers NTSA to identify any defects and take cars of the road.
I saw one last question and then maybe we’ll turn it over to the next panel.
SPEAKER: Thank you. You said that autonomous vehicles replicate our
ability to see and think. And I would argue that they also will hear and remember. So --
MR. KAN: They what?
SPEAKER: Hear and remember. And so I think that in the priorities that
you listed I didn’t hear data privacy and security issues as one of them. And I understand
that there are a lot of people currently claiming jurisdiction over those issues and it is not
necessarily a place where DOT has had to tread. But at what point do you see DOT
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
93
needing to engage with data privacy and security issues when you’re engaging with this new
technology? I realize I’m asking for another prediction, but.
MR. KAN: The question was how does DOT think about data privacy?
Because one of the components of what cars do is not only see and think, but also hear and
remember. And that’s absolutely true. The reason self-driving cars are getting better is
because they have an algorithm that learns from the past, similar to how we drive.
The role of data privacy is a very open question which has tons of different
views. I have a personal view and sense of where things are going, but it’s very unformed
because there are two major issues here.
One is there is not a clear legal authority for DOT to acquire, regulate, track,
mandate any type of data standard, and that presents a complexity here. There are pieces
of law we can interpret, such as different types of authorities we can interpret broadly. But
there’s not been any clear direction from Congress what the desire is for DOT to do or any
federal agency to do around data privacy.
However, there’s a very good model out there. It’s called the ASIS model,
A-S-I-S. It’s a great model; people should look. After I forget if it was ’93 or ’94, there was a
summer in which there were a number of significant passenger plane accidents in this
country. ValuJet was one of them in the Everglades, New York had a second one. And
what air carriers did is they came together and said, you know what, we should figure out a
way to share information in a way that we don’t trip any anti-competitive laws, we protect our
trade secrets, and we don’t get into trouble with these things. And so they created a system
called ASIS where twice a year for I think two days the top operators from air carriers share
data.
I was landing at JFK and my warning indicator started beeping, I don’t know
why. Another person in the room says the same thing happened to me. And suddenly,
through pattern recognition different air carriers with different pilots started pattern
recognizing and understanding some of the problems they ran into.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
94
This is not -- there’s no congressional mandate, there’s no legal requirement
to do this. But what Congress did do is it passed an indemnity provision that said if you
share data for public good or public safety for this specific program, you’re indemnified from
prosecution provided it wasn’t intentional and provided you report within 24 hours. So
there’s basically a voluntary data-sharing organization that came about.
And I think this is going to be one of the important ways that this is
addressed, that as AV automakers start coming together and sharing data, that it’ll address
some of it. But it doesn’t address the other side of that coin, which is personal data. And
that’s one that there are tons of different views. My personal prediction for this, again
channeling Yogi Berra, take it for what it’s worth, is I think you will see Europe much more
aggressively thinking about this than the United States. With GDPR you’re already seeing
question around data, around what tech firms are tracking.
And so what tells me is you’re going to see other countries move faster on
this data privacy piece I think than we are simply because that’s an issue that they think a lot
about. Whereas here, it’s unclear, is this a state issue? Is it a federal issue? Where will we
derive our legal authority from? And what are the mandates that we put on them?
So what I see in this country at least there have been two modern attempts
to create a voluntary data exchange. The problem with that is when you have companies
that are at very different life stages, pure startup to companies that have invested $50 billion,
there is no desire to share data. And so what we’re seeing with some of the resources, as
some of the panelists are going to talk to next, is what should they even be tracking? And
my hope is before we get into even privacy, we begin to just align on what is the specific
safety standards we want data to report on? And let’s start sharing those things before we
start getting into the very sticky wicket of insurance, of data privacy, and data sharing.
So with that, I very much appreciate Korok and Brookings for hosting this
fascinating discussion. I hope you all stick around for the next two panels and really
appreciate your engagement on this issue. It’s really active citizens, engineers,
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
95
policymakers that together are going to help shape this technology that our kids will look
back at and say it transformed the world.
So thank you very much. I really appreciate the time. (Applause)
MR. RAY: Okay. Thank you. So we are going to move forward with our
next panel on Policy.
This will be a little different from our morning sessions -- so that timer is
something I should get. The next time I get into an argument with my wife I'm going to bring
that.
So, we are going to have more of an open-ended kind of broader
conversation about policy. I would like to try to connect our conversation to everything we've
done so far this morning, the technical topics as well as the keynote. So, basically what
we're going to try to have as much of an open conversation as we can, and to address many
of the issues that have come up.
So again, just to level set, I think this is -- I think we're all on the same page.
We want to understand how this technology can eventually develop. And I think it is a big
challenge that will require the coordination, or at least the cooperat ion of multiple different
parties.
To me this area is wide open, there's still -- you know, we're in the very early
innings of the game, and as a community I think we need to think through how exactly this
will work out with both private sector, public sector, and everyone in between.
So let me introduce our panelists who will speak about these issues. W hat
I'll do is, I'll do a quick intro, they'll kind of do a longer intro on themselves, it's always optimal
that way. And then they will give a little bit of an opening statement to talk for a few minutes
about their thoughts broadly on this topic, and then beyond that I'll go into some leading
questions, and if the discussion is not conversational enough I'll try to be more provocative.
Okay. So first I'd like to introduce Cliff Winston. Cliff has been at Brookings
since 1984, he was the -- at one point the Editor of the Brookings Papers on Economic
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
96
Activity. Also in a prior life he was Associate Professor at MIT in Civil Engineering. He has
a lot of writing on autonomous vehicles, and also on privatization and deregulation. He's
had many books and essays with colorful titles like, Let's Deregulate All the Lawyers. He
has a PhD from UC Berkeley just like everyone on the A&M infrastructure panel.
Okay. I'd also like to introduce Cliff -- I'm sorry, I just did Cliff -- Rick. Rick
is a Professor of Policy Analysis and Management at Cornell. He's the Founding Director of
the Cornell Program in Infrastructure Policy, CPIP. So, he is a real expert on infrastructure
policy. He was at the Council of Economic Advisors three years before I was, so 2004 to
2005, and he has a PhD from my favorite university, University of Chicago; second-favorite,
A&M is my favorite now.
Marjorie is a Senior Policy Researcher at RAND, she has worked at the
Office of Science and Technology Policy, and she was also an Associate Provost at
Georgetown. She has written extensively on autonomous vehicle safety and policy and has
an MPP from Harvard.
So sorry to repeat the credentials, it must be my Indian American upbringing
to constantly think about my credentials, that's how I was raised. (Laughter)
So that's all I'm going to say up front, and then I'm going to sit down, and
then why don't we just go sequentially from my -- from left to right on my side, just to talk a
little bit about -- say a few more remarks about yourself and about your views on
autonomous vehicles, and then we'll get more specific. Does that work?
MR. RAY: Yes. Five minutes timing.
(Recess)
MR. WINSTON: It's been a while. Two articles were -- Sunday on
autonomous vehicles, in the paper, I'll go through those briefly to frame my comments. It's
not to point out it was fake news, to point out it was incomplete news which I'll fill in from the
perspective of transportation economics.
The first one was in The Times, and it was one of these ones it says: We're
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
97
not going to get autonomous vehicles tomorrow, they'll be long off. What we're really
missing in that piece was some historical perspective. It really would have been helpful to
mention the airline industry. The 1920s even though the industry was growing there were
many people who did not think we would ever have a commercially viable airline industry,
even in the '30s when the industry was moving along people said, aircraft technology has
peaked, we're not going to see better planes than what we've got now.
Obviously they were wrong. The same kinds of problems are going to exist
with autonomous vehicles.
What would have also been helpful was to point out the development of the
airline industry was not hampered by its own progress but by the government . We learned
that very quickly when we had regulation and deregulation in the '70s, and we saw what that
did, constraining pricing, entry/exit, and so on and so forth.
And remarkably it exists today. We still don't have open skies on all our
routes and most importantly we don't have cabotage, foreign carriers cannot serve our
routes. That would be handy now given that Boeing has grounded its 737 Max, al l that
could have been filled in with foreign carriers.
The other problem is infrastructure, we've still got public airports, still got
public air traffic control, you can't get any more efficient than either of -- inefficient than either
of those and those have hurt the industry.
The second article was in The Post, also on Sunday. That was one: What
are the failures we've learned about autos that we can learn to try to improve autonomous
vehicles? Maybe you saw that article. I didn't get any of the failures out of there, I couldn't
figure out what they were talking about.
The New Yorker has a similar one, equally bad. It's very obvious what the
failures were, again all from highway policy. All the inefficiencies that have compromised
automobile travel are going to make autonomous vehicle travel even worse, all right.
It hurts the transportation sector, but importantly, which wasn’t mentioned
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
98
earlier, it helps the -- it hurts the growth of the economy.
Okay. So my views on the industry are positive and optimistic when it
comes to the private sector for an obvious reason, it's incredibly competitive. W e've got U.S.
automakers, foreign automakers, technology companies, and most importantly there are
huge stakes.
This is not the case of GM making Chevy Volt, and if it doesn't work out, you
know, they'll make money with SUVs. If it doesn't work out, they are through, all right, and
they cannot count on 20 years: President Ivanka Trump or President Alexandria Ocasio-
Cortez bailing them out, it isn't going to happen. So all of them know this, this kind of
competition is exactly what we want, all right.
So, where are the problems in the industry? It's from public policy, all right,
from the get-go. First and foremost, why didn't we have testing passed? What was the
problem? Why couldn't Congress pass AV START? Why didn't NHTSA step up? They
could have done it and got a national framework for testing, all right. It really would be nice if
policymakers would get behind this industry and recognize what it could do.
The second, infrastructure: this is the public, automakers don't own the
infrastructure, the public sector has really got to handle all of this, the city and state, and
ultimately the national level, if they don't do anything that's going to ground these vehicles in
their place.
Finally, most importantly, what inefficient highway policy has done, all the
way down the line, inefficient pricing, inefficient investment, inflated production costs,
regulatory delays, misallocation of funding from states, and cities, and so on and so forth,
and most importantly, constraints on technological advance and innovation.
All of that, huge welfare losses for non-autonomous vehicles, and will do
exactly the same to autonomous vehicles. That is what you've got to improve, all right.
And if you don't think this is important for the economy. Suppose Trump
says, you know, I've really had it with the West Coast and Washington State, let's ground all
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
99
of Boeing's planes, let's really see what's up with them. They'll never vote for me anyway.
What do you think that's going to do the economy?
It won't just be delays in the travel sector, the whole economy will be
affected, labor markets, production, trade, so on and so forth, so when you get
improvements in delays, which is what autonomous vehicles can do, it affects the whole
economy.
We estimate about a 1 percentage point increase in GDP growth from
autonomous vehicles, that's what's at stake.
I will conclude with a positive note. The good thing about all of this is that
everybody knows it. That is: the public sector is going to be competing, cities, states and
government, the national governments, and they already are. Right?
Bloomberg has a website that says: here are the cities where autonomous
vehicles are being tested. They're going to have another one, to say: here is where
autonomous vehicles are operating. As word gets out, and some cities and states are
readying up for doing this, there will be a political cost to this.
So I'm hoping, in the same way that competition is really going to help this
industry develop, no question in my mind about that, it will also put pressure on
governments, expose the political costs, and hopefully billion dollar bills, or even trillion
dollar bills won't be lying on the sidewalk.
MS. BLUMENTHAL: So many ideas, so little time, so, much to respond to.
I wanted to frame some starting comments based on a handful of questions, but I think I
should also address Korok's zeroth order question, which is to observe that I have a long
background, a career in science and technology policy, I've come relatively recently to
autonomous vehicle safety, but that is the area in which I have been focusing when it comes
to AVs.
That said, I can draw from looking at the rise and the varied applications of
information technologies across the board, including how policy has and hasn't worked in
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
100
those areas.
So, my first question is: What is safety? And one of the things that
surprised me in my recent work is there is no consensus definition, there's no consensus
definition in general, and I found that out in talking to people at the National Safety Council,
as well as elsewhere, and there's no consensus definition within transportation.
So, at some level we don't really know what we're talking about. In our work
we define safety as avoiding harm to people, whether they're in the vehicle, or on the
roadway ecosystem.
You heard about crash worthiness from Derek Kan, that's one angle, and of
course -- it was interesting he didn't mention there's also occupant protection, and that's
another major strand of the Federal Safety Standards.
But if we look at the last decades of the previous century, you know, we've
gone from 1950s and seatbelts being very rare, but beginning to appear, to debating about
different passive restraints in the 1970s, then the '80s the fuel economy drove us to reduce
vehicle weight, which of course added to the safety problems,
And steadily through the end of the last century we started to see more
electronics in vehicles which is what brings us to where we are today.
Where and how can AV be measured: is my second question, and that is
one that that we looked at in depth, I'm not going to go into depth at the moment, but it is
important and again Derek Kan touched on this, to recognize that many things affect AV
safety, beginning with the components.
And so it makes a difference if you're AV has cameras, radars, LIDARs or
everything since, as we just heard, that that sensor stack can be quite varied among
vendors. Testing and measurement are immature, and I would say to my panel member on
the left that that's one reason why we've seen some Federal policy slowness, let's say,
because we don't yet know enough how to do those things very well.
You heard earlier that people talk about how many miles the vehicle
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
101
traveled. I have colleagues who estimated that a vehicle would have to drive hundreds of
millions if not billions of miles for statistical confidence of their safety. That's not happening.
It's certainly not consistent with what we're seeing with the 1,400 vehicles you just heard
were being tested.
We are doing a lot of simulation, and simulation is very helpful, but as
anybody who really understands simulation knows, you can only simulate what you know to
be able to simulate. And so there are going to be situations that are rare but important and
sometimes fatal that will not be captured in simulations.
Finally, AV developers are not telling us what they know, and that makes it
hard for policy makers to plan for safety, part of that is because of the competition you just
heard about. This is a great race, and people who are in that race don't want to share.
Third question: Can we compare AV safety to safety of conventional cars?
Well, that's what everybody wants to do. We all know conventional cars, we want to make
the comparison, that's obvious, but a computer system on wheels does work differently from
the combination of a human being operating something on wheels that has computers and
other stuff in it.
The other comment I would make on the challenge of comparing is that it
may be easier to avoid being a cause of an incident as opposed to being a victim of one,
and that's probably true for both AVs and conventional vehicles. If you imagine a three-lane
highway, and if you are the vehicle in the middle with three vehicles to your left, three
vehicles your right, one behind you, one in front of you, if one of those vehicles screws up
there's not much that you can do about it.
Fourth question: How much safety is good enough? And that relates to the
question: Is it reasonable to expect that AV's be at least as safe as human-driven vehicles?
Safety advocates either want them to be at least as safe, some of them
actually want them to be better. I have colleagues who did and analysis that showed that
once you get to that, at-least-as-safe point, then there are benefits in terms of lives saved
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
102
and other problems avoided by getting those as safe vehicles out on the road. So you
should beware having the perfect be the enemy of the good.
And then my fifth is a grab bag of: What are some other important
considerations in the context for AV safety? First, AVs are not consistent in their design or
operation, they operate within specified sets of circumstances.
And you heard that, again, from Derek Kan talking about a thousand miles
across the country and whether you do or do not encounter other vehicles or pedestrians
along that way.
We do expect the fleet to be mixed for a long time, and that's consistent with
the slowness to getting meaningful deployment. There is a big difference in terms of safety
between whether you have a minority of vehicles in the total fleet that are automated, and
then the majority are conventional cars driven by people, as opposed to some anticipated
future where most, if not all cars are automated. The whole world works differently in that
second scenario than in the first.
We do expect -- again you just heard a little bit about that -- there to be
more communications between vehicles, between vehicles and infrastructure, between
vehicles and people, and so on, but that all isn't going to happen for a long time. When it
happens it will help with safety.
So, finally, I would say that it's important to think about safety from a
systems perspective, and I learned that when I looked at some of the work involved with
what's called vision zero, the quest to eliminate all traffic fatalities ideally by the middle of the
century.
If you're going to do that you're going to change your vehicles, you're going
to change your infrastructure, you're going to change your policy, you're going to change
your incentives, everything will change; and so as we think about what it takes to make
safety we can't focus on the AV alone, we have to think about all these factors interacting
together.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
103
MR. GEDDES: Thank you. So, I'm not going to spend any more time on --
I'll just get my clock going here -- on bio, because I'm going to try to solve the world's most
pressing policy problem, and offer a suggestion related to autonomous vehicles, and do it in
about five minutes. So that's all I've got.
So you heard Derek Kan talk about the megatrend, and the megatrend was
mass urbanization, and along with it increasing traffic congestion. So, as the world
middleclass grows people, first they want to get a car, they want to move into the cities, and
they of course, you know, drive those cars in increasingly dense environments creating
traffic congestion around the world.
There's a whole slew of social problems associated with that, there's wasted
fuel, there's wasted time, there's fine particulates produced from diesel fumes that have
shown that babies who grow up near increased diesel fumes have worse health outcomes. I
could go on. But that is going to be, and is, a huge and increasing social problem.
Now to the economist, the economist who looks at traffic congestion, Econ-
101 says: every bit of traffic congestion is due to the mispricing of the use of the road at that
time. It's due to the mispricing of the use of the road at that time. That is, the price of the
road is too low. If you could get the price of the road correct you could regulate the market
both on the supply side, and on the demand side in order to make the flow of traffic through
that facility smooth, so basically going at free flow.
So, imagine we could do that. Imagine we could, and the panel of
engineers I think agreed, that driverless cars are a way of getting people to accept road
usage charges per mile a little bit better than they have in the past, they're a tool.
So imagine we could do that, we could price per mile, but that price per mile
is not fixed, that price per mile is variable, okay. Like, like the price of bread, the price of
beer, the price of ties, I don't know, those prices vary but they vary in real time, and the price
per use of the road at that particular time varies.
And it varies in a way to clear the market, so when an economist says clear
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
104
the market that means that the demand for road space exactly equates to the supply, and
the cars flow through at maximum cars per hour.
Now, you could imagine that market for road space clearing centrally so that
there's some -- what we call an independent system operator -- an ISO perhaps that serves
as the central clearing mechanism for that market, okay.
If that seems, I don't know, pie in the sky as we sit here, the market for
wholesale electricity, probably the electricity we're using in this room right now, the
wholesale market not the retail market, clears down to the second in a way to equate the
supply and demand for electricity on the grid.
The same is true for radio spectrum. If you went to Chicago the market for
number 10 wheat would be clearing that way right now. The market for pork bellies, the
market for coffee, the market for orange juice, all these commodities are clearing in this, you
know, sort of constant way.
So, imagine now that we had that data, so we knew we were pricing not
only according to per mile but in a variable way, in a way that clears the market, and then
that price data were plugged into your app, okay, and that that price was the key thing that
say your Google Maps, or your Waze, whatever app you want to do, was telling you how to
get to work, okay.
So you say, I'm going to use an autonomous vehicle, I've got to go to work, I
hit the button: how much is my trip to work in the cost right now? Well, according to current
road prices it's going to cost 6,075 cents.
However, there's a there's a little arrow, and the little arrow tells you whether
the price is going down or whether the prices going up. So, if the price is going down you
wait 15 minutes, you answer some emails at home, you have a cheaper road price to get to
work.
If the road price is going up you say, I've got to get in the car right now, give
me that driverless car, I'm going to work right now, okay. So, that is, imagine both, not just
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
105
the market currently, which is what we call a spot market, but also a forward market which is
what you have in almost all commodities.
But the market is a new market, it's the market for the use of road space,
and that forward market if you're worried about price risk, allows you to hedge if you're a big
user, say like United Parcel Service, DHL, FedEx, you want to lock in those prices, you
might hedge in the forward market for road space.
So that is the basic. I think the key thing, we've talked about safety, critical
issue, we've talked about potential time savings, less aggravation associated with driverless
cars, but my hypothesis, my strong hypothesis first is that we could eliminate all -- I'm using
that word carefully -- all traffic congestion through accurate prices of roads that would clear
down to the second or every few seconds, and we would have a spot and a forward market.
And second, with leading to that, okay, to eliminate one of humanity's most
pressing problems, in my view as an economist, would be the single greatest benefit of
autonomous vehicles. If we could get people to think about the trip differently, and the way
they pay for roads differently that would be a huge advantage.
So, if you want to read about this some more, I'm honored to have a paper
in the journal Nature, the British journal Nature with two esteemed economists Peter
Cramton and Axel Ockenfels. If you want to read it give me your card that says: send me
the Nature paper, and we have a longer -- that's a comment almost -- we have a longer
paper in the academic journal called Journal of Institutional and Theoretical Economics that
works all this out in gory detail about how the Independent System Operator would work.
Peter Cramton has set up electricity -- wholesale electricity markets around
the world, he's also set up spectrum markets and, you know, we explained how this would
apply to the market for road space.
So as I think about this as an economist, I love safety, I think it's important, I
think there's a lot of challenges, but I would hope you think about autonomous vehicles as
leading to this new economic world where we accurately price the use of roads in real time.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
106
And I'll stop there.
MR. RAY: Okay. Thank you, panelists. So, let me just begin with sort of a
broad question. This morning we saw that there are some -- there are a lot of opportunities
and challenges with -- on the technical side of autonomous driving, and then Derek Kan said
that we're a little bit slower on the timeline than he expected.
Can you, all three of you maybe, just speculate on how do we get to a path
towards adoption both on the private side, in industry, as well as on the public side? In a
sort of a thoughtful and proactive way, how would we -- how would that happen? Go ahead.
(Off-the-record discussion)
MR. WINSTON: They're not independent, right?
MR. RAY: Right.
MR. WINSTON: I mean, the private sector -- and that was my point, the
private sector is constrained of what the public sector is doing, right. They can't just --
there's sort of this Tesla fantasy. They can't just make cars and just dump them out there
and say, go adopt them. I mean this is not how non-autonomous cars worked, right. You
realize that.
The way we have cars out here is they still have to go through a Federal
regulatory inspection, right, for safety -- you know, for safety reasons, and of course and
there's obviously licensing, you know, and all that.
So, you know, this is not going to be something they can do by themselves,
they're going to try to work obviously as quickly as they can subject to competition, but the
government really has to take the lead, first getting the testing framework nationally.
I mean, you know, this notion, Marjory, that Mitch McConnell and Elaine
Chao held up the legislation because of their concerns about AVs doesn’t make any sense.
If they were so concerned, why are they allowing the cities to do it? We have testing going
on all the time now, so they haven't stopped it.
I mean they could have taken the lead and they didn't. It wasn't one of
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
107
McConnell's priorities, and I really do not know what Elaine Chao cares about. But the truth
is, that's what happened, and it's a fantasy to think otherwise.
So that's where that has to happen. They need to take the lead to get that
going, then once that happens then, as we know, this is going to be a cooperative activity,
there's going to be a lot of learning back on that RAND Study that says we need to go X-
trillion miles, which is ridiculous, and it's a terrible study, there's learning by doing, right.
You're not just going to have to go a trillion miles, each time you drive you're
going to learn a lot. So that obviously is going to cut down the amount of time you need
before these things will get you get well.
So the point being is that then we have that secondary time when they're
testing and they're learning, there's obviously going to be some cooperation there. I'm not
quite sure how long that will take, you know, it depends who emerges as a leader, right, and
how much politics will go in.
You know there are going to be low-end people that are going to try to slow
things up, so that they're not going to lose, and then there's the final stage, so then the
transition to broader adoption, and then the policies that facilitate efficient usage.
Again, all of this involves heavy involvement of the public sector, and you
know the history, it's not great about what they do in terms of constraining efficiency,
innovation and technological change.
MR. RAY: Marjory, do you want to answer that? (Laughter)
MS. BLUMENTHAL: So, I actually was going to go in a different direction,
which I will get to. But I think what we just heard is that a reason we don't have a national
framework is that we do have a situation where people are learning by doing, not quite let a
thousand flowers bloom, but we do actually benefit by having multiple experiments, in
multiple tart parts of the country, by multiple government entities, multiple companies, and
so on.
And because they are all leading us in different ways nationally and
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
108
internationally, we are learning by doing, and although I haven't dwelled on the subject, I
think that that actually provides a basis for not jumping in with a Federal intervention that
might be poorly informed and poorly formed. That's my response to my colleague to my left.
But I think in terms of how might we move forward? It is interesting, you
know, that there were some references by Derek Kan to liability, and liability is driving many
people to assume that when these vehicles come on the road they're likely to be deployed
as parts of fleets.
You know, we can talk about Tesla all we want, which is making an
individual consumer retail play, but we're talking about vehicles where, if that sensor stack is
not well maintained then there is a risk that the vehicle will not perform as it is supposed to
do, and if there is no human driver then the only people who are going to be blamed are the
owners, operators, developers of the fleet.
They have a lot of incentive to get it right, and I think that is how we're going
to see some of the initial rollout.
MR. GEDDES: So, great. So to review, the question is how do we get
adoption?
MR. RAY: Yes.
MR. GEDDES: And now we go back to Cliff's point that all the infrastructure
assets that are discussed here are publicly owned, the states are owned entirely -- the
interstate highway system is entirely owned by states. I don't think there's a foot of it
anymore that's owned by the Federal Government.
There're city streets, there're county roads, all these assets are owned by
some public owner. If there's -- not owned by anybody let's get some free infrastructure, but
of course they are. So you have to deal with public owners, so one thing that concerns me
about these discussions is it was alluded to on the last engineering panel a little bit, is about
the public ownership of the infrastructure assets.
So to get any of this stuff deployed in a serious way you have to take their
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
109
interests, and their incentives, their budget constraints, et cetera, into account.
So, you know, running the Infrastructure Policy Center for the past 10 years
at Cornell, you know, I love these people. They're all wonderful, they are risk averse. If
you're the state highway official the worst thing for you, is to see that something about the
highway system on the front page in the newspaper tomorrow, you bear all the downside but
you don't bear the upside.
They're budget constrained, so if you go with them to a proposal to spend
money it's probably going to, you know, go in one ear and out the other.
So what, I'm going to launch a term that I would like everybody to keep in
mind, it's called "value capture", okay, I don't know how many people have heard of value
capture -- but you've heard of it. It's a great -- I know Baruch has heard of it. It's a great
term that is gaining traction in infrastructure policy.
So the idea is that there's enormous value in everything we're talking about,
that the value is latent in autonomous vehicles, there's safety value, maybe congestion
reduction, et cetera. The question is: how do you share that value with public sector officials
in a way that they benefit from the adoption?
If I have time for a quick example, from my own town of Ithaca, New York, in
Upstate, we had a hundred-year -- and I love sewage treatment plants that's what I do -- we
have a-hundred-year-old sewage treatment plant, you know, the same thing with the big
settling pools, and skimmers, et cetera, tremendous methane dispersion into the
atmosphere, methane is a bad greenhouse gas.
Johnson Controls comes in and says, hey, folks we will install a methane
digester, which is a giant white sphere larger than this room, that will capture the methane
as it's processed, we will use that methane to turn three or four, it's called micro turbines that
turn at 120,000 RPMs.
Those micro turbines will produce electricity for this plant, enough to cover
the whole cost of the electricity for operating this plant, and we will install that technology to
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
110
you, the city of Ithaca, at absolutely no cost.
The way we pay for it, is allow us to capture some of your reduced electricity
bill. Allow us to capture some of your reduced electricity bill, we'll bond against that and
raise the money to install the technology.
I got a tour of the plant, it's so successful the city of Ithaca Sewage
Treatment Plant is selling electricity outside the plant onto the grid. Johnson Controls, or
somebody's going to install another methane digester.
Its neighboring counties are actually bringing waste to be processed, paying
the City of Ithaca to do that processing, and I believe the water discharged in the Cayuga
Lake is actually cleaner than it was before.
So, I urge us to think about applying that model to driverless cars. One
super valuable thing are the streetlight poles, right, the poles themselves you can put up
cameras, 5G, et cetera, you install LED lights instead of pressurized sodium that will give
you more lumens at a lower electricity cost, just share some of that with the public sector
owner so that they gain and say, by the way, we want to, you know, help us with the
autonomous vehicle thing.
So think about those win-win situations that don't require the public sectors
to spend money, and are low risk, and I think that's -- that it's part of the solution.
MR. RAY: Thank you, Rick. So, I will actually want to follow up on that
point. We've heard a lot about infrastructure today, and you're an expert on infrastructure
financing and -- tell us a little bit more about how smart infrastructure in particular could
develop in the future under new and novel financing arrangements that were possibly not
possible under the kind of -- the older model? What new opportunities does this, all of this
new technology bring on the infrastructure side?
MR. GEDDES: Yes. So, great question. I mean I think that a lot of old
structures should be relied upon first and that's -- one of the things I've learned as a, you
know, sort of observer and commentator on this area is that everything in this sector is risky.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
111
Any time -- just think about building a bridge okay, kind of a common thing, that everything
about that is risky.
There's design risk in the bridge, there's construction risk, there's
geotechnical risk, there's strike risk, there's operating risk, there's acts of God, force majeure
risk, you know, I could go on down the line.
And so the companies involved in this, you know, risk is like a hot potato,
they want to get rid of it, right, and so they're trying to distance themselves or control the risk
as much as they can so a disaster with the infrastructure doesn't bring down the parent
company.
So what they do -- this is lawyer intensive -- is they form SPV, special-
purpose vehicles, which are these legally created entities that manage the risks, they issue
debt and equity themselves, they are the actual legally contracting entity. So, to deploy this
new -- trust me this is all risky -- what we've talked -- everything we talked about today is
risky, to deploy this and that's -- the key thing is that the debt issued by the special purpose
vehicle is issued by revenues coming into the SPV only, so that's called project financing.
So we would use project financing structures particularly. One of the
interesting things I'll just -- just a footnote -- is bond life. In other words the bond market
usually has 5, 10, 15, 20, you know, very fixed-term bonds, and I think that matching the
bond life to the payback is getting more flexibility in that bond market, whether it's taxes on
municipal bonds, or taxable corporate bonds, is something we could do to help facilitate this.
But what I would do is go back look at -- you know, there's a whole world of
finance of this, infrastructure financed on Wall Street and elsewhere, take that expertise and
think about how to modify existing project financing models to adapt to the new technology.
MR. RAY: Yes. Cliff, go ahead.
MR. WINSTON: So, so an interesting thing historically, in this country and
probably everywhere else, its modes, virtually all of them lead infrastructure. We first had
cars before we had roads, we first had planes before we had airports, and air traffic control,
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
112
and so on. And that is how I think it's important to think about how autonomous vehicle
infrastructure will evolve.
We're going to first have the vehicles. They're going to continue to get
better before we get the technology and infrastructure to accommodate them. Now, what's
going to be interesting though is that effectively the pressure that autonomous vehicles are
going to put on infrastructure providers.
Normally people tend to think about financing which is not unreasonable,
however, there's to be something else they're going to have to think about, and that is the
performance of their infrastructure, because if it performs badly it'll significantly compromise
autonomous vehicles' performance. If it performs well it'll enhance autonomous vehicles'
technology performance.
Now, there'll be a number of ways, both economic and technological, that
this infrastructure can succeed. As I noted, just messing up on pricing and investment in
things that should have been done with non-autonomous vehicles but were never done cost
us quite a bit, but that's masked over the fact that we spend so much money to cover it up
and have trillions of dollars invested. That really will show up with autonomous vehicles.
We will have induced demand, so if they're concerned about congestion and
you don't introduce congestion pricing that's obviously going to compromise to a significant
extent what autonomous vehicles can do.
If you do not get build the roads properly and you have a lot of potholes
autonomous vehicles aren't going to like that, and that's going to really hurt their
performance as they try to figure out what to do.
If you're going to blow the chance to optimize on capacity, and now making
lanes thinner or narrower which you now can do instead of making them wider, anticipating
there'll be free flow traffic that never materialize, that's going to waste a lot of capacity.
That's just on the economic side.
On the technological side that's going to be harder because you're then
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
113
going to have to start interacting with the vehicle and say, okay, at various levels, at the
street level, optimizing stoplights, things like that.
The question I asked earlier in the day was, basically getting at the point, we
have not changed traffic like technology for decades, they are based on the assumed traffic
flows that exist. So when you show up at 2:00 a.m. and see a red light there's no one
around, and you're wondering: why isn't this thing turning green or blinking red, you can
thank the decades-old policy that is still used to do that.
Obviously that's the kind of thing we want to be changing with autonomous
vehicles. Then more broadly trying to coordinate then massive movement of autonomous
vehicles on highways, and so on and so forth, that now is open for a lot of different
explorations.
Utah is partnering with the private sector entity, Colorado is partnering with
a private sector entity, all trying to look at best ways to do this kind of thing. But I think
ultimately of the autonomous vehicles getting out there, they're going to put the push on the
infrastructure to develop and then the competition will exist, and we will see what best
practices are and we'll see the costs and benefits of who succeeds.
MS. BLUMENTHAL: I would just add that picking up on the point I ended
on about safety being a system, a lot of what's going to happen with AVs is going to fit into a
larger system of activities. So, land-use planning is going to change. You know, some of
that pressure on road use that Rick had talked about may change with AVs because they
can go in sort of bizarre routes and park themselves, and do things that we don't expect to
see with people, and cities are going to plan for that.
Cities are in fact, as you've just heard, working on their traffic management
systems and between them and the interactions with AVs we now have new cybersecurity
concerns. That's another risk that is going to drive investment, because if you don't attend
to that you're going to have huge problems. So I think systemically we'll see a number of
motivators for investment and for creative responses.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
114
MR. RAY: So let's talk more about safety, and it's given that Derek and Ken
both said it's kind of the number one priority, and you mentioned this before. So, we're now
in a world where say 40,000 fatalities a year on U.S. highways, and there's one crash in
Arizona and Uber shuts down their program.
If there is such an emphasis on safety how are we going to get there if all
technology is very preliminary, and there's going to be some growing pains, how do we --
how do we overcome that when -- even if there's -- and there's a long-term benefit, but
there's going to be short-term costs. And how do we -- how do we as a society get there?
MS. BLUMENTHAL: Well, you heard Derek Kan that projections have
changed for when we're going to see significant deployment, real commercial deployment.
And I think that that is a reflection of industry recognizing that the problem is harder than
they thought it was. And so having this internalized adaptation is going to be critical.
I think because of the competition that I mentioned before among
developers, we have yet to see a true sense of shared fate that is it wasn't: oh, Uber had the
problem therefore I'm clean.
Everybody suffers, public trust, and confidence suffers when there's any
incident, so one of the things that my report had recommended is that when there is an
incident, because it is so uncommon. that people share information.
And again you heard Derek Kan talking about an aviation example, but the
more people in the business share information at least among themselves, arguably also
with the government, the more we will learn from these scarce incidents, and be able to
progress together. But the most important phenomenon that we see now is that people
have lowered the temperature of their enthusiasm and the hype, and I think people are self-
moderating.
MR. GEDDES: Can I just add?
MR. RAY: Yes, go for it.
MR. GEDDES: So, to me -- this is a great, great question, Korok. So, to
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
115
me it's analogous to what I tell my policy students at Cornell about type one versus type two
error at the FDA. Where type one error is you release a drug that actually harms people,
type two error is where you don't release a drug that actually helps people.
MR. RAY: Right.
MR. GEDDES: And you see the type one error, but the type two errors like
an opportunity cost that you don't see. So, I think part of the role of policy people is to
emphasize the cost of not having a safer autonomous vehicle.
So, you know, the weird things that I calculate everybody -- many people up
here probably have been to the Vietnam War Memorial, and I look at all the names on that
wall, the terrible loss, it takes us about 18.23 months to kill that number of people on U.S.
roads and highways, okay.
So that visualizes it, they're disproportionately young people, just horrible
carnage on the roads today. So every week, every day we don't have this new technology
deployed, all these young people, those crosses you see on the highways we're being lost
because of human -- the majority is human error.
So I think the jobs of economists is to hammer on opportunity cost, use the
type one versus type two error, you know, analogy and just say, look, yes we know someone
was killed in Arizona, it's terrible, it's horrible, but we have to -- this is better than the
massive, you know, people drinking, and people texting, and everything else that's causing
this carnage on the roads. And I'll just keep stressing that.
MR. WINSTON: So, when Derek was mentioning his example, I don't know
if any of you market-oriented people were thinking the same way I was, but he's talking
about airlines landing and one of them having a problem and saying, you know, why can't I
share that with all my fellow pilots, where was the FAA? When did the FAA take note of
that?
What if the FAA says, hey, we'll keep track of all of this when you have a
problem, and we are in charge of promoting safety in this industry, and we will make it aware
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
116
to all of you, you know, when a pilot has a problem at one of our airports, and then hopefully
that will help induce cooperation? Where were they?
Why do we have to have the private sector say, hey, why don't we organize
and have a cooperative effort where we solve this problem? I'll give you my answer in a
second, but this is how things work when you're dealing with the government. I mean these
are things that are occurring even now, already there have been coalitions of various
automakers and technology companies, now, in autonomous vehicles to promote safety,
okay.
Whereas, again, and I don't know why this hasn't gotten across, the Federal
Government has to actually authorize the standards for autonomous vehicles for them to be
sold, they can't unless that happens.
Congress authorizes NHTSA to do it, that's what we need the legislation for,
then they can sell cars. They can't do it before that, all right. They have to take the lead
that's just the law, okay.
And that's really what we should be looking for but we're not getting that
kind of leadership, all right. Ultimately it's going to come in pushing from the private sector
to do it, and they're frustrated, because they really would like the public sector to engage,
but they don't.
Why? Status quo bias, if you look repeatedly, decade after decade, policies
that should have changed long ago they don't, and then when you have technology being
such a challenge as it is, these guys don't know what to do, and that's what -- this is not
new, this has been going on forever, and it's probably going to continue to go on forever.
It's going to be an uphill fight, but hopefully through these demonstrations,
and again this is a global effort, so it will get attention in this country too, we'll get
somewhere, but it's going to be a lot harder than it has to be. Defend the government.
(Laughter)
MS. BLUMENTHAL: I think -- if I recall correctly, Brookings has published a
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
117
lot of work talking about regulatory capture, and the fact is in a lot of industries where there
is regulation the government can never understand the world as well as the people that it is
trying to regulate.
SPEAKER: True.
MS. BLUMENTHAL: I cannot forecast well enough what will happen in this
space but I have seen and read enough about what happens when regulatory policy goes
wrong to understand that if the whole system, the industry, the governments at different
levels, if everybody goes a little slower maybe we have more of a chance of getting it right.
MR. RAY: So I have one more question for the panel, and then we'll open it
up. And the question is about sort of long-term cost and benefits from autonomous
transportation, in particular for, say, the labor market and the capital market.
And so what are some things maybe we haven't discussed yet of benefits or
costs of autonomous technology, say for all of the truck drivers in the U.S. today, or for the
price of land across the country. Can you either speculate on that or offer any opinions?
MR. WINSTON: Okay. So I actually will answer that, but let me fill that in.
You know, I may -- I told you that there's an estimate of GDP growth. Okay. The intuition,
I'll give you one intuition, Boeing -- let me go back further. Let's go back to the Stone Age
when we did not have a transportation system, okay, so all your production was, you know,
where you lived in your cave. You know, you worked where you lived in your cave, you
know, and so on and so forth. Okay.
What did transportation do? What it did is enable you to get all the benefits
of modern economy, economies of scale, you can produce and sell to many places, product
diversity, you didn't have to eat the same -- I know what you ate then -- bison or something.
You know, competition, diversity in labor, you know, all these markets
opened up, so you can think of infinite congestion in the limit, that's the Stone Age, you can't
move, all right. So as you get incremental improvements in congestion basically what you're
doing is opening up all the things in an economy that we get from a transportation system,
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
118
and that's what autonomous vehicles will do.
Not just the transportation system, but it's going to open up all these
sectors. Now, admittedly we're not where we were in the Stone Age, but we obviously suffer
from pretty extreme congestion that affects everything, labor, trade, competition, and so on
and so forth.
So the hope with autonomous vehicles operating efficiently with the right
technology and the right economic policy is we will get these congestion improvements that'll
make the whole economy come alive even more.
Now, it's disruptive. Again, governments don't like it, labor markets some do
-- some parts do some parts don't. Obviously labor gets hurt when there's capital labor
substitution, okay. So the truck drivers obviously in this -- in the seat, they're out of work, all
right, and that's the offsetting part, or the negative part that people get nervous about, it's
disruptive.
Okay. But there are two other parts that offset that, that are very important.
One, productivity increases, output goes up more. So even taking the technology as
constant, you are at least going to do more, and you need at least some labor, you're not
completely substituting away from all labor to produce that output.
But here's the big-ticket item. New occupations, all kinds of things will open
up with new technologies, jobs that we didn't have before, new things we didn't have before,
and what's important is with the growth of the tech companies in the tech economy most of
the new jobs or a large fraction of the new jobs are in new classifications of occupation.
So you put that all together and you're saying: yes, there will be some loss
in employment from capital labor substitution, but productivity increases and the creation of
new jobs are quite likely to offset that, and obviously there are going to have to be the
adjustments, as there always are with a new technology.
MR. RAY: Okay.
MR. GEDDES: So, I want to address that. I agree with everything Cliff
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
119
said, Cliff has done more analysis on this, but just kind of a broader -- a broader view. If you
think about transportation in general, you know, it's been treated almost like a utility, and the
thinking about it. So nobody should go without clean water, nobody should go without
electricity, once we have that nobody, you know, should go without a number of other --
maybe communications, nowadays nobody should go without heat and, you know, that's just
I think a great part of American society as we sort of believe these basic needs should be
met.
And that's been true for transportation for a long time. So, in the early part
of the 1900s there was something called the Farmers to Markets Movement, and this was
the idea that farmers had all these crops that they were growing more than they could ever
eat themselves, and they wanted to get it to market.
But a lot of the roads were dirt roads so when it rained they couldn't move
their crops to market, so there was this rural road movement to try to get the farmers to allow
them, you know, to get their crops to market.
And then the very name of the interstate highway system is to connect the
nation together, sort of the old postal service notion of binding the nation together, except it
was done through a -- you know, a system of interstate highways.
So as we, you know, move to this new technology; I guess I'll just tell
another quick story. Everybody here who is in D.C. in January should go to the
Transportation Research Board Meetings, the TRB, and I love that meeting, right. And they
have an exhibit hall there, they have all cool -- they had the autonomous car last January.
So early January they had a car moving back and forth, you know, I think you could ride in it.
But one time I went there, and it was a couple years ago, and there were
these guys, kind of interesting looking guys, standing next to a vehicle that was on one of
the earlier slides which was an autonomous bus maybe or six- or eight-seater, right. And,
you know, he said this is interesting, but (inaudible). I said, oh. They said that we're running
this around Boston, right, and we're running it -- I said, you know, really that's really cool.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
120
And I said, but doesn't that -- there's an MTA in Boston, don't they have
exclusive rights to operate buses in Boston -- yeah I'm a policy guy right -- so how do you
how do you overcome this? And they said, oh, we're running our autonomous bus where
the MTA in Boston doesn't serve. I'm not indicted them, it just doesn't serve these
underserved areas.
And he explained to me the way the app work. You have an app and they
say this autonomous bus -- you are two blocks from this autonomous bus stop, five other
people want to ride this bus from A to B, come to this corner in six minutes, and this
autonomous bus will take you from A to B, okay, at a very low cost.
And so that's how they were doing it. So there are these gaps, we don't
think any neighborhood should be cut off from some form of transportation. There are these
gaps in our system that I think autonomous transportation is filling, right.
So I see that. Now, how does that get back to your jobs question? A key
thing is access to jobs. It's that, can people get to jobs? In New York City where I spend a
lot of time, it's really expensive to own a car, right, so you have to have some form of public
transportation to get to a job, and I think that if we, you know, creatively use autonomous
vehicles, they can help solve that age-old transportation problem of really providing universal
service. And that's just one example.
MR. RAY: Okay. Marjory did you want to say (inaudible).
MS. BLUMENTHAL: I would just add that when it comes to -- when it
comes to trucks we benefit again from the fact that technology is not changing as fast as
people think. So if you talk to people who are in the trucking industry they talk about various
assists, and they talk about first assistance in the long haul portions, as opposed to the, you
know, getting on and off of a highway, and so on.
There are also people who will talk about how the age structure of the
commercial driving population is relatively high because those jobs are not attractive to
younger people. So there can be problems getting enough people to do today's trucking
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
121
jobs.
If you add that together with slower than some people would like to see
deployment, you may have a self-correcting problem, which is not to say that some people
will not lose work and certainly opportunities will be foreclosed, but as the new kinds of
opportunities that Cliff mentioned come online, again the system may regulate itself and I
end up being a little bit less worried about trucking than I do about other kinds of jobs.
MR. RAY: Okay. Let's open it up to questions. I know people might be
nervous about asking the first question, so let's go straight to the second question.
(Laughter)
SPEAKER: Thank you. Actually I would like to ask three questions in one
minute if it's possible. Marjory Blumenthal talked about the AV timing. Aren't we asking too
much of AVs without providing a lot of support from all of us? Because the assumption is
AVs are super intelligent machines looking for zombies on the unorganized asphalt, when
we have urban planning that could organize the urban mobility by speed and by modality.
And we have psychologists, and educators that could educate the population in terms of
behavior.
The second question for Rick, if I may: land value capture, and public
goods, and public value. My overarching principle is if you have an agency that provides a
public good it should have an endowment such as land-grant universities, in this country's
long tradition, such as foundations, such as metro that is land, public transit is one of those.
So you can use land value capture like Tokyo and Hong Kong, in order to provide that
additional support.
The third question to Clifford Winston. Mayor Bowser wants to spend a ton
of money to buy nine electrical buses. They would be -- no, no, I have a lot of respect for
the Mayor, this is my personal -- but that would be expensive to buy, it would be expensive
to operate, and logistically difficult. Wouldn't it be better to use a pilot on those grid corridors
and have something like Navya, you know, pilot the shuttles, electrical NAVs. Thank you
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
122
very much.
MR. RAY: Yeah, just to give everyone a chance. Maybe just answer one of
those questions, so just one of you, to pick. Yeah.
MS. BLUMENTHAL: I think in response to your question, when I was
talking about a systems approach, or referring in response to something that Rick had said
to land-use planning. I agree that there are many kinds of planning, many kinds of
personnel that have to complement the vehicles themselves. It's never just the vehicles
themselves.
And, you know, we heard earlier about public awareness, even some of that
work perhaps, it's the psychologist you were referring to, to advise the public on what it is
they can expect. You know, there was a case in Boston where a consulting group got
together with the city government, I think it was, and had an AV petting zoo.
I mean that's just a way to, you know, get people exposed because you
have to pay money to get into TRB, and that we need to help the rest of us.
MR. RAY: We have to be maximizing?
MS. BLUMENTHAL: Yes. Yes.
MR. RAY: Yeah, let's just keep it quick so we've got the --
MR. WINSTON: Pretty quick.
MR. RAY: Yes.
MR. WINSTON: I'll alienate the rest of you. Public transit was an
experiment in this country, okay. Initially we had private transit that was before public, all
right. It's been a mistake economically as it turned out, it's an experiment that failed. From
an economic perspective if you look at the user benefits and even account for any savings
and externalities, and congestion reductions, and compare that with the cost and now the
huge and ever-growing subsidies they get, they are not socially desirable; all right.
And the advent of network transportation companies like Uber and Lyft is
just putting the final nail in the coffin. Transit was just a mistake, it will be one of the benefits
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
123
of autonomous vehicles is it will make it increasingly clear that those systems should be
eliminated, and that we will have effectively personal transit companies.
So I am not supportive of any spending on transit, if anything I think we
need to realize that subsidizing people like myself, you know, from coming in from the
suburbs really just has not worked out, and that the mobility that autonomous vehicles wil l
get us will give us the final push to get rid of those systems.
MR. RAY: Okay. Let's take the next question -- let's see -- how about right
there? Yeah, you can stand up so we can -- yeah.
SPEAKER: Hi. I'm Eta Nahapetian from Fairfax County Virginia, and we
are working on trying to get an autonomous shuttle happening over the next few months in
Fairfax, and we're looking -- and we're focusing on making it a transit solution, because we
think the opposite, and that we're hoping that these AVs become -- help us get to a transit
solution so we don't have this crazy increase in traffic. As you might know Fairfax has a lot
of traffic.
So we're really struggling with some of the regulatory issues. And going
through figuring out how we're going to work through the NHTSA waiver process. We're
partnering with Dominion to -- they're going to help purchase our vehicle but we're -- as
we're trying to jockey to figure out what kind of vehicle we're going to get, we are struggling
with the NHTSA waiver process.
And I guess maybe this was a question for Daniel -- for Derek Kan, but do
you know if there's any kind of movement to not make it such a black box? You know, it is a
black box to local government to whether -- what the regulatory requirements will be.
MS. BLUMENTHAL: The simple answer is. I don't know. But that sounds
like a question for the Conference of Mayors or League of Cities who I know are looking at
some of these issues.
MR. RAY: Okay, great. Now, our next question up here in the front; yeah,
great --
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
124
SPEAKER: Thank you I had a question about the safety elements of
autonomous vehicles. If you look at sort of everything the USDOT is saying, and what a lot
of the OEMs are saying: safety first, safety is our top priority, presumably from the OEMs
that's because they see a consumer desire to hear that the vehicles are safe. But then like
Rick was saying earlier, is we already accept the vast amount of fatality on the road as
consumers.
And so there's something about our personal utility choice that we decide to
make these trips when we know the statistics and we know how dangerous it is. So how do
you reconcile that moving forward? What's the barrier there from an -- I guess economics
perspective, which I don't know anything about? And how do we push past that?
MR. GEDDES: I mean I guess, to go back to the FDA, I'm old enough to
remember the AIDS crisis and, you know, there was the AIDS Cocktail and the -- you know,
it wasn't 100 percent solution, but the FDA was holding up the deployment of those drugs.
And the gay community got together, and they lobbied, and they demonstrated, and they
said, look, you know, some of us have AIDS and this drug is available, you know, let us have
it, right.
So I would take -- I mean it's a heroic analogy, but I would take that -- and
again I think economists are not vocal enough in pointing out opportunity costs. This is
something we cover in the first week of Econ-101, right, it's the things you do see, you know,
which are the people who die from the AV in Arizona, and the things you don't see.
The things you don't see are the people who would be alive today if they
had had an AV, instead of the human error that's killing, you know, 30,000 people every
year. And so I think the role of the policy people, and the economists in particular, is to be
much more vocal about the opportunity costs.
And to say, yes, fatalities are terrible but, you know, in the deployment of
any -- I mean aviation was like that early on, it wasn't long after the Wright brothers flew, that
the first person was killed in an airplane crash.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
125
So, you know, I'm not minimizing that harm, but I'm saying we have to
stress the harm to society of not having the technology deployed, and these regulatory
holdups are killing people. Yeah, I could give you other analogies, but you get the point.
MR. WINSTON: What you're getting at though that's important, is the notion
what we call heterogeneity. You see there isn't this "we", we're all different. In other words,
when you look at the breakdown of auto fatalities what do you see? You know, drunk
driving, right, not wearing a seatbelt, texting, it 's what we call heterogeneity.
All right, I'm going thrash the engineers for a second. There was a question
about what major one should have to do this stuff? A little econometrics wouldn't hurt.
This is what the Engineering Safety works, it's just awful. They go out and
they get police accident reports, right, and they fit models to what we call severity, all right,
and they use the data on the police accidents reports: Were they drinking? Were they
speeding? Was there a hairpin curve? You know, all that kind of stuff, right, and they draw
results from that.
Well, the obviously dumb thing about that is this self-selection, right, you
only -- you're in that sample if you got into an accident, right. It's not a random sample, what
you'd want is a sample that has everybody, and most of those people don't get in accidents,
because they take actions, right, to prevent that.
That's what autonomous vehicles are doing, they're eliminating the
heterogeneity, all the dumb decisions that people have made to self-select, and put
themselves out in a snowstorm, you know, when they run into somebody. You know, or
after they've been drinking they run into somebody, or run off the road, autonomous vehicles
are going to say, sorry, you can be that way but once you get in us you're all the same, all
right.
And that's really how we're going to get the improvements, it is we're cutting
the heterogeneity making everyone the same.
MR. RAY: Okay. Let's -- yeah, go ahead.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
126
MR. GOLOVIN: Karl Golovin. I'm wondering if a defining moment will come
when apparently Tesla is working to set up its own insurance company for Tesla owners as
well as its own collision repair because, you know, their cars are unique enough you don't
want to take him to a -- you know, a traditional repair shop.
So once the owners of Teslas; and Tesla itself, through its insurance
company, is willing to accept the liability that the assurance that they believe their self-
driving vehicles are sufficiently safe, they're willing to incur the liability; won't that be a
moment when all -- many Teslas are already on the road will become autonomous?
MR. WINSTON: They really need Federal approval for all this (inaudible),
they can put anything they want in them; they cannot sell them and operate them.
MS. BLUMENTHAL: I also don't think that a model of complete vertical
integration, where a company does everything and self-insures is scalable. So, it would be
interesting to see what happens, but we've heard a lot of announcements and
pronouncements coming out of that company; and then reality, you know, looks a little bit
different.
SPEAKER: Yeah, exactly.
MS. BLUMENTHAL: So I'm not going to worry about that one.
MR. RAY: Okay let's take one in the front, yeah?
SPEAKER: It seems like this -- I call this NEAT system, Network Electric
Automated Transport. It seems like these this whole NEAT System is based on having --
you know, a $700 mobile smartphone with a 40 to $80 a month data plan in order to be able
to use it.
Is there a way -- you know, do you know -- you see the future being able to
use a Hale System -- or a transport like this, without such a device in your pocket? Or it's
assuming that you have to pay this expensive data plan and have this, you know, $400
iPhone thing in your pocket in order to use, you know, this network transport? That's my
question. Thanks.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
127
MS. BLUMENTHAL: Right now there are a number of models that people
are experimenting with for different kinds of platforms, supporting mobility as a service.
Some of them draw on public information as well as private information. This is an evolving
space, and AVs are a part of it. My guess is that just as all of that kind of matching, and
planning, and payment services, as that's evolving will also end up with different approaches
to accessing it.
So if there is a digital divide issue, which is what I think you're getting at,
then we'll also be looking at some alternative public kiosks, or other options. They may not
be as flexible as having a device in your pocket, but ideally, for reasons that have been
covered by others in government, local government will make that kind of thing happen. And
we may end up with different kinds of technology innovation that will lower the cost of the
communications and information aspects.
MR. RAY: Let's take the last question. Yeah, go ahead.
SPEAKER: All right. Thank.
SPEAKER: I'm sorry. A comment to Cliff, and a question for Rick; and I
usually agree with Cliff but I think that -- I think the FAA was very involved in the ASIAS
Program. I know they were very involved in developing a sister program which encouraged
pilots to share information on near-miss accidents. There were so few accidents beginning -
- around that time that you needed to start to extract information from near misses. And so
the FAA encouraged pilots to do that, and gave them the same sort of indemnification that
the ASIAS Program. So I would give FA credit for that one.
My question for Rick; Derek Kan said the Federal Government was neutral
between internal combustion engines and EVs, this was in response to a question, internal
combustion engines create carbon emissions that's another unpriced externality. Do you
agree that the Federal Government should be neutral in the AV debate between electric
vehicles and -- or hydrogen fuel cell and internal combustion?
MR. GEDDES: Do you mean, so you're asking my own personal view on --
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
128
SPEAKER: Yes. Well, your policy view as an economist. Yeah.
MR. GEDDES: Well, so I'm sort of in favor of electrification. I mean I think
there's -- a lot of reasons to reduce fossil fuel usage, carbon dioxide is one of those, there's
others. One thing -- and I think there's this sense that autonomous vehicles and electric
vehicles are kind of converging, so just you know --
SPEAKER: Yes.
MR. GEDDES: Again, to underscore my earlier point, I hope that the
takeaway from my earlier comments is that everybody thinking about this will think about
real-time road pricing as the third piece, right. So autonomous vehicles, electric and road
pricing is kind of -- kind of converging, right, so I am not fuel agnostic and, you know, it
would take a little bit longer to talk through that, I can think of a number of reasons.
One thing though that I'll just say, that is a cautionary note and, you know,
one of the great things about me as an economist studying infrastructures, I get to talk to
engineers, a lot more different part of campus, and they're very concerned about
infrastructure resilience, particularly in the State of New York after Hurricane Sandy. That
was a big thing for our state.
And one of the things you do and, I'll just toss it out there. When you rely
more on electric vehicles you integrate the transportation system with the electric system
more profoundly. So if you were to -- if somebody were to take down the grid and you can't
get a charge on your vehicle, that's a real problem, right.
And what we study hurricane after hurricanes, one of the biggest problems
is actually -- the impact of the hurricane on infrastructure itself is big, but people can't get
fuel after the hurricane. The pumps themselves at gas stations run on electricity, right. So,
one of the problems, biggest problems, biggest impacts, is the days after the Hurricanes
where people can't get fuel.
So, I guess I'm just cautioning to think about the impact of integrating your
transportation system more deeply with your power grid, because your power grid then
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
129
should be much more -- stronger, resilient to either, natural disasters, terrorist attacks, cyber
events, et cetera. So, I'll just toss it out there. But I personally am not fuel agnostic, I think
there's reasons to favor electrification.
MR. RAY: Okay. We are out of time. So let's respect everyone's time.
Thank you to the panelists. If you have further questions you can come up. (Applause)
MS. SNYDER: All right, we are going to get started now. Thank you all,
first of all, for being here for this really excit ing conversation about how autonomous vehicles
are going to change the world and the ways that we need to change the world to
accommodate them; and what role we want them to play in our world.
We’re going to be talking about public policy. Policymakers have struggled
to develop a framework for brining autonomous vehicles onto our roads. To catch you up
quickly, the House of Representatives overwhelmingly passed a bill in 2017 which got
bogged down in the Senate last year and died there.
The Department of Transportation, as you heard from Derek Kan, has
published three versions of voluntary guidance, but no binding regulations. So, that’s kind of
the environment that we find ourselves in right now.
I’m Tanya Snyder, I’m a transportation reporter for Politico Pro. Hopefully
you all subscribe to Politico Pro Transportation, or at least read morning transportation, and
if you don’t, come talk to me afterwards (laughter).
I want to introduce the panel, too. To my left is Tom Madrecki, Vice
President of Supply Chain and Logistics for the Grocery Manufacturers’ Association; Beth
Kigel, Vice President at HNTB, and Director of their Smart and Connected Solutions
program; Baruch Feigenbaum, Assistant Director of Transportation Policy at the Reason
Foundation; and Jyot Chadha from the New Urban Mobility Alliance at the World Resources
Institutes Ross Center for Sustainable Cities. She leads NUMO’s work to develop
partnerships on tech and mobility.
So, thank you all for being here, and thank you all for being here. I want to
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
130
start with this question of not just the ways that we need to transform our cities and places
for autonomous vehicles, but the way that we can put them to work to make our cities and
places the best cities that they can be.
We hear a lot of potential benefits that seem to solve a lot of the most
vexing issues that cities have about congestion, emissions, safety, access, mobility; how can
cities best put these to work so that, you know, we’re not just, sort of, making space to
accommodate what industry wants to throw at us, but that we are using them for our own
purposes. And we can start with Beth on that, actually.
MS. KIGEL: Can you hear me now? Okay, great, so it helps to be flexible.
Well, one of the things that I’ll share with you about my background is, prior to coming to
HNTB, I was a Florida Transportation Commissioner for seven and a half years, and I led a
chamber of commerce at the same time for about seven years.
So I come at this from a little bit of a different perspective; kind of, a
combination of policy, technology, and then understanding community engagement, and
economic development, and quality of life issues.
So, first and foremost, I would say heard a lot about -- we all here today,
we've been educated quite a bit, I think, on the things that we've heard. Think about some
of the localities out there and how much they do not know and understand. In terms of what
cities can do, cities first and foremost, can start to really become educated on the benefits
and the risks of what is happening with autonomous and connected technology.
The other thing I would say, is to be open to actually bring pilots into your
city. I think that’s critically important because not only is it something that might help the AV
entities, but it also helps you learn more about your city; where the pitfalls are, and things of
that nature. Where do you need to improve?
Miami and Ford had a really great relationship when Ford came in with their
AV testing down in Miami, as you might expect a very congested environment, just an
interesting place to test. So, great information for all to get.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
131
I would also say that collaboration is very, very important as a city, and even
as a county, between private and public sector entities of all kinds.
I led a chamber in Palm Beach County. We actually were able to immerse
all of our entities in an education about autonomous connected vehicles, as well as smart-
city technologies. When you start bringing together not only elected officials and
administrators from cities, not only private sector participants, including hospitals and others;
you start bringing all of them together, along with the Urban League and what have you, you
really start to get an education about how these advancements can really advance a locality
and be better for a quality of like, and, I heard earlier, competitiveness, because we are in a
global market and how our quality of life stacks up, and what kind of efficiencies we have
within our regions and our localities, the more competitive we’re going to be as well.
MS. SNYDER: Great, whoever wants, take it next.
MR. MADRECKI: I think, just building on what you said, I think that one of
the really important things to also consider as a city, you need to proactively consider the
future deployment of AV technology is that this isn’t occurring, or isn’t going to occur in a
vacuum.
And so, there are many different things happening all at the same, right.
So, if you have just the tremendous, sort of, growth and demand for ecommerce delivery, for
example. Prior to my role at GMA, I was the Director of Urban Innovation and Mobility at
UPS, and so I would talk to cities all the time about the tremendous influx of demand that
curb space, and the strain of that physical real estate asset in a city, and how that demand
shift is happening.
You have the emergence of micromobility, you have the emergence of TNC,
all of these things are happening, and so you can't just say, let's take the AV bucket, or let's
just take the TNC development; all of this is happening at once. And so, as a city, I think
you need to adopt this very integrated, proactive view.
MR. FEIGENBAUM: I would say, I think one of the questions is what about
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
132
cars? Because, obviously, cities have got concerns about congestion and about mobility ,
and as was mentioned, about the curb space because there's going to be a lot of
competition.
And so, our approach at Reason, not surprisingly as a free-market
approach, has been pricing. And I know it was mentioned a little bit on the past panel, but I
think there's a lot you can do with pricing.
If you're worried about zero-occupant vehicles, sort of, the zombie cars, as
they are sometimes called, roaming around, then you can price those vehicles higher. If you
want to encourage carpooling, you can price those vehicles lower or not at all.
Now, we don’t have a great history for carpooling in this country . There's a
lot of hope, and Alain Kornhauser at Princeton has spoken on this, for those of you that
familiar with him, that with some of the automated vehicles, and the improvements with
ridesharing, and what we see even today with smartphones, there's going to be a lot more of
that.
We also have to talk about the 800-lb gorilla, which is transit. And
obviously, in some metro areas, New York, D.C., others, transit as we have it now works
okay. I’m not going to go quite as far as Cliff went and say it’s a total disaster (laughter), but
it needs a lot of improvement in a lot of places and it’s not great.
And I think automated vehicles can do a lot of that in terms of maybe getting
rid of the labor costs down the road, looking at smaller vehicles, right -sizing. I think
partnerships with the private sector is going to be important as well, working with Uber and
Lyft for some of the low-density routes. We see that already with first mile and last mile.
So, I think that pricing for cars and figuring out how we can make transit
better are two important things.
MS. CHADHA: I think I’ll focus a little bit on what Tom brought up, which is
on micromobility, and I’ll just take a quick poll: how many of us have been on a shared
scooter or a shared cycle? Okay, that’s a lot.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
133
The point of view that I think I’m coming from is that today we have that
opportunity to start flexing our muscle, and start to have discussions and pilot an experiment
on the topics that will be important when AVs are eventually on our roads, today itself with
micromobility.
So some of the topics that are hot with micromobility right now is the use of
space, and the allocation of space, is pricing, as you were saying, and thinking about what
are the different models behind that; around regulations, around permits, thinking about
market entry, and I think we really do have that opportunity today to start linking some of that
groundwork for when AVs are actually on our streets.
Where micromobility and AVs differ, though, I think, is that in the current
ecosystem that we have today, AVs could very well fit into the space that we have for
personal cars. And I think that we do have the opportunity today; cities and other
stakeholders, do need to come together and thing about how do we use this moment in time
to really think about the goals for our cities, and what sort of cities we are trying to live in.
So, we would really push for cities that are active and are shared in terms of
mobility; so, thinking about a 20-minute neighborhood. We think that, clearly, land use and
planning play a really big part of making that happen, but a lot of these new mobility entrants
are starting to add to the possibilities in that space.
So I really think that engaging on the topics that are at hand today gives us
that platform to be able to get our cities into a position that we have the right sort of
infrastructure, we have the right public/private partnerships, we have the right mode choices
that we would want for different types of purposes and different types of populations.
MS. SNYDER: Picking up there a little bit, how as we think about remaking
our infrastructure for a new autonomous era, how do we avoid some of the mistakes of the
highway, where highways were built through low-income neighborhoods, where they
accelerated sprawl; how to we make sure that we are working sustainability and social
equity into that infrastructure?
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
134
MS. KIGEL: Going back to the point that I said, you have to start looking at
that data, and the statistics, and putting them at the forefront, to really bring that to reality.
One of the things that I would talk about, since I am from Florida and I
served on the commission, Florida actually is embarking on its Florida Transportation Plan,
and they're putting some of those statistics right front and center, with all of the stakeholders
that are looking at it.
So, looking at how many people live within a half a mile of a healthy food
source, looking at how many live below the poverty line? How many households do not
have a bank account? Actually, 8.4 million households in our country do not have a bank
account.
Transportation in Florida is the third highest household cost for a family.
56.5% of our residents live within a half a mile of fixed-route transit. So that means the rest
of them don’t. So, we’re actually putting those issues at the forefront and having those
discussions so that is part of our overall transportation and smart-city planning.
Social equity is becoming more and more of a hot topic. I mean, it’s a
national discussion relative to this and others. So, the more leadership that’s taken within
our communities and our states, and etc. on this, the more we’ll include those in our
solutions.
MR. FEIGENBAUM: I guess I will just say, to be a little contrary in here,
after the first answer, we have to realize, obviously the highway building era had some real
problems in terms of building through neighborhoods. I mean, there's a situation in Atlanta
where there's a highway that actually takes a curve when going straight would have been
the cheaper and more direct route, because they were trying to separate the low-income
minority from the central business district.
That’s a pretty awful policy, and we certainly want to find ways that we don’t
do that in anything, really, again. And I think looking at the overall costs, you know, in some
ways, the Federal government can actually be a problem here, because to the extent that it
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
135
is incentivizing certain policies.
When it was incentivizing highway construction with a large Federal share,
cities and states were sort of doing what the policy designed them to do, which was build
roadways even if it wasn’t necessarily to logical places, or it wasn’t the cheapest place,
because there was that incentive there. And so, we have to make sure we’re tapping the
right Federal incentives.
But I will say, we don’t want to forget that highways and roads in general
move the mast majority of folks. Certainly, lower in cities, and certainly lower in D.C. and
New York than at other places around the country, but even if the automated vehicles are
shared, they are still going to be on roadways, and we have to make sure we have sufficient
capacity for that.
MS. CHADHA: I think to add to these points, I'd also say that -- and I’m
sure that this has been covered in depth through the day, I’m sorry I couldn’t be here earlier
-- but that we have the opportunity today to rethink what a vehicle looks like as well, and
many of you in the room are engaged with that.
And so, we have that choice to think about wheelchair accessible, service
animals, how are we going to build vehicles to allow for that level of access for all. But I
think the point also goes back towards, is there political will, is there general will, what is our
mindset around these topics.
MR. MADRECKI: And on the subject of political will, I think that -- and I’m
going to go a little philosophical in this, but I think that part of demonstrating political will is
also the ability to step back and to say that you don’t make infrastructure equitable; you start
with equity, and then you would sort of underline, or sort of think through, well what is it that
you even want to achieve in the first place, or do you need that infrastructure to make that
connection.
You start with that equity goal, and then your finding ways of moving people
around, rather than just be, “Well, we need to build a bunch of roads because we need AVs.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
136
Oh, by the way, we need to incorporate equity as a part of that.” I think it’s just like a
philosophical reorientation.
MS. SNYDER: I think that’s what I was trying to get at with that first
question, also, of, what are the problems that we’re trying to solve and we figure that out
first, and then figure out how AVs can help solve those problems; rather than say, they’re
coming! What do we do!
I wanted to ask, maybe at the risk of fixating too much on Tesla, you know,
Tesla has had several high-profile fatalities. Uber, obviously, their self-driving car that hit a
pedestrian last March; how have these very high-profile -- and understanding that we have
37,000 deaths on our roadways every year, but these get a lot of attention -- how do these
incidents change or impact the approach that cities or states are taking to regulating or
embracing driverless technology? Start with you, Beth.
MS. KIGEL: I want to look at this first from a little bit of a different angle
because we heard earlier today about how the OEMs are responding to this as well. I think
what this is doing is, this is raising significant awareness that this is not as basic and as
simple as many thought.
I mean, when you're trying to replicate the human mind and the decisions
that are being made that way, it’s very complicated. And so, actually, in talking to some of
these OEMs, they have pulled back and they have really intensified their testing processes.
Some of them have their own tracks, in fact, and they are putting them through much more
rigorous testing.
And, one of the things that I think is recognized, we actually did a national
survey that dealt with a number of items, including public acceptance. We found out that
52% of the people that we surveyed with the national sample poll actually bel ieved that the
greatest benefit from all of this is increased mobility for non-drivers. That’s the number one
benefit, and then safety is second.
But how the public actually responds, whatever their first experience is with
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
137
an autonomous vehicle, or any of these other modes that we’re talking about, is going to
really shape things. So, just like we heard earlier, if there is a fatality with an AV that’s being
tested on a road, that’s going to affect the whole industry. There's pullback on that side.
I think we’re seeing a little bit more patience within our communities, and on
the public sector side, an appreciation of that.
MR. FEIGENBAUM: I would say, to directly answer your question, I hope
they don’t because I know it’s easy to look at something like the Uber crash, and say, “Boy,
this is a dangerous technology and we shouldn’t do it.” But as was pointed out in the last
panel, 37,000 lives a year is a lot of people that die and even if automated vehicles are able
to reduce that by 50%, that is still a tremendous improvement.
Now, we all hear the statistic, I guess it’s 94% of crashes are caused by
humans, and the reality is that while a lot of those will be gotten rid of, probably not all of
them, and AVs may introduce some new crashes. So, I don’t want to be too Pollyannic here
on the options, but I do think a 50% improvement is better.
And I do think it would be helpful if there was some Federal -- the
legislation, the Self-Drive Act, I happen to think it was a good bill; it wasn’t perfect.
Now, I can say there were folks on both sides of the spectrum; folks who
thought it went too far, and also someone who’s worked on AVs for a while, who’s name I
will withhold, that said it thought it was an abomination. That being said, passing legislation
is very challenging, and I think it will be a good start because the Federal government is the
one that has traditionally done with safety, whereas the states do with licensing. And I think
if we introduce new dynamics where states and local governments are specifically looking at
the safety of vehicles, I think that’s problematic for a lot of reasons, including the number of
different modifications that the OEMs would need to make to vehicles.
MS. CHADHA: I think that just adding to that good point that you were
making, you look at the number of deaths and accidents that have taken place with AVs --
you can count them on one hand -- you look at the number that are caused when you're on
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
138
a cycle or on a scooter, and then you sort of, like, in scooters you're probably getting to
around the tens, and then cycles and cars and pedestrians goes up a lot higher; and it’s
always interesting to see the uproar that’s caused by the one or the two, versus just the
complacency with the every day.
But aside from that, I think that it’s also interesting -- I can't speak for Tesla,
but I think it’s interesting with some of these other new mobility companies like Uber and
Lyft, about how they're becoming, not a right-hailing company, but a platform company,
right. Aside just from the new modes that they're bringing into their platform, and the new
modes that they're trying to invent. It’s also the one-stop shop for you to get your
information and make your travel decisions, to make your purchases for transport as well.
And I think that when companies start thinking in that sort of holistic way
about the entire transport ecosystem, that starts to position them -- or I’m hoping that starts
to position them -- differently in the minds of people, rather than honing in on the, like that
one accident that took place. Not to minimize any death, but I think that sort of thing is taken
out of context, out of proportion.
MS. SNYDER: I want to ask also, and Baruch, we’ll start with you because I
want to ask about Federal regulation, obviously a lot of people think that the bad PR that
goes along with a fatality is enough incentive for OEMs, for AV developers, to wait until the
technology is ready. And like you said, when is it ready? It’s an open question because
even if they're going to improve things, but it’s not, you know, they haven’t worked out all the
kinks if they are going to make it 50% safer, maybe it’s ready?
Putting that aside, a lot of people also look at it and say, when we’re looking
at corporate accountability, we also have Takata airbags, we also have VW emissions
cheating, can we rely solely on corporate self-policing? Is there a role for the Federal
government to sign off on these vehicles before -- and I realize that this is something that no
one in the Federal government seems to have much appetite for, but I think that when we
see the kind of reservations that the public has about this, maybe it’s a conversation worth
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
139
having.
MR. FEIGENBAUM: Sure, yeah, I think that’s a very good question.
Couple of thoughts, there are some -- obviously the state of California requires certain things
that the auto makers have to do, and many of them are in fact testing there. And there is the
self-certification platform where the auto makers do have to show that their vehicles are
testing, and the Federal government does have to approve that. So, there's a little bit of
that, and should there be more?
I guess I would say not at this time, and there are a couple of reasons for
that. First of all, the Uber crash was awful, and I think nobody was more upset with it than
me. And the reason I was upset with it is because I really believe that these companies
have a responsibility to put safe products on the road, and clearly, that’s not what was going
on. This Uber vehicle was clearly not ready for primetime.
The good news is I think the market, and also the state of Arizona, punished
them for that action, that I think was honestly far more punishment than they would have
gotten from the Federal government if that was the only thing they could have gotten.
And the other challenge goes back to the part of your question, which is,
does the Federal government know when these vehicles are safe enough, how do you prove
that, how do you weight the cost benefit of potentially a loss of life here versus getting the
car on the road sooner and doing the loss of life there; I don't know. It’s hard to say.
I don’t have confidence that the government would have enough information
to actually weigh that correctly.
MS. SNYDER: No one wants to argue with Baruch on this (laughter)? All
right, the Senate is about to markup their five-year reauthorization bill and getting back to
infrastructure, Derek Kan said infrastructure is not an enabler, but potentially a blocker; that
it is something that can get in the way of the deployment of autonomous vehicles -- though it
can’t necessarily, kind of, conjure them.
If we’re about to set policy for the next five years of highway policy in the
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
140
United States, what should the Senate, what should Congress be looking at now that in the
next five years will need to happen?
MS. KIGEL: Well, they might be able to look at what the state of Florida is
doing, in fact (laughter). I know I keep going back to that, and I promised I would (laughter).
MS. SNYDER: It’s okay, I keep going back to Tesla.
MS. KIGEL: But, actually, if you take a look at what Florida is doing now;
legislation has been enacted and projects are being sought to actually embed the
technology in new infrastructure. So, when we’re looking at long-range planning, this is
something that should be a part of that.
We’ve talked a lot about the life-saving elements of this. Well, if we’re going
to be serious about it, we’re going to start including these things in our long-range planning,
including things that have to do with the electrification of vehicles, and expecting those on
our roadways, for example.
So, in looking at long-range transportation plans, how do you address that,
and if that’s the way the automobiles are going for various reasons, then you have to ensure
that the infrastructure can support that, you have to find ways to start being forward-thinking
about how do you handle things such as hurricanes, that we heard before under those
circumstances. So, these are some things that can be looked at in transportation planning.
Also, we had a discussion about 5.9 Gig that happened earlier, what kind of
leadership role can the Senate play in that to ensure safety over the long-term, and what
have you.
MR. FEIGENBAUM: I would say there are currently grant programs right
now for a lot of the smart infrastructure that I know Beth has been talking about, and I think
those are important, and I think those should continue, that’s obviously separate from the
five-year authorization bill, but I still think that’s important.
I also think there's value in the so-called dumb infrastructure, which is
maintaining your roadways, making sure your lines are painted. And Kirk Steudle, who was
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
141
former Michigan DOT Secretary, brought up a good point, which is that, well, if you're in a
cold weather place, your lines and your roadway pavement can get chipped pretty badly in
the winter.
So, I think we need to look at some sort of partnership, possibly a
public/private partnership, where state DOTs are going out and making sure that the
maintenance for the self-driving vehicles is actually there. And that’s not necessarily just the
pavement, but it could also be if there's vehicle-to-vehicle; I’m personally more of a fan of
cellular v-to-x, than I am of DSRC, I think that’s kind of, well, we’ll see where that goes, but I
think that is really important and that has the advantage of, not only helping automated
vehicles, but also helping today’s vehicles and transit vehicles, and so shouldn’t be
challenging from a political perspective.
MR. MADRECKI: Just to add a little bit about targeting dumb infrastructure,
and taking more of a nuts-and-bolts approach, but I do think that there's a degree to which
that you can sort of write in some degree of prioritization about specific -- it could be a freight
pinch point, it could be particular areas where spending money matters.
And so, a lot of states have done this in terms of right-sizing infrastructure
projects, and thinking about, well what is currently in the pipeline and do you actually need
the 90 million dollar project, or could you accomplish 98% of it with a 9 million dollar project.
So, I think there's ways of sort building in that philosophy of prioritization
and then right-sizing into how the actual reauthorization is structured.
MS. CHADHA: I’ll just pick up on the point that Beth was making about
hurricanes. And I am wondering what the opportunity is to look at funding resilient
infrastructure in cities and marrying that with the recreation of transport infrastructure that’s
needed as well.
And I think that there is common cause with a lot of these new mobility
platform companies that are looking at streets that are slower, that are looking at re-thinking
street allocation and I think that bringing some of these moving parts together can actually
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
142
spread that forward.
MS. SNYDER: I've been sitting here trying to think whether I want to go
down the rabbit hole of 5.9 gigahertz that you brought up Baruch, let's go there for a minute.
The FCC is still kind of in this decision-making process about whether to
reserve this kind of safety band for vehicle-to-vehicle or vehicle-to-infrastructure
transportation communications. it’s very underused right now.
There's also this kind of raging debate about what is the right -- maybe it’s
not even raging anymore, maybe it’s settled -- that the original DSRC technology that the
Federal government had, sort of, chosen, is not the right one. DOT now wants to be tech-
neutral and aside from that, most of industry has embrace the new technology.
So, right now DOT has just chosen to stay out of it. They have kind of
frozen rulemaking on that. Is that the most helpful thing for them to do, to just stop. Or, do
they need to, at some point, just say, this is the policy, and move forward. Because it does
seem like this limbo is more harmful than having new policy.
MR. FEIGENBAUM: The answer to your question is, yes. For the time
being I would say -- because I don’t think it’s settled -- that they should stay out of it; but yes,
at a certain point in time -- and, of course, I wish I could tell you when that time is, and I don’t
know -- we do need to choose a technology because we can't just go on this forever.
I think the rationale for staying out of it right now is that as it stands
currently, DSRC, would be more effective, because 5G v-to-x is not ready for primetime. But
many people expect that within a few years, 5G with v-to-x is going to be the better
technology, but because, of course, we can't predict the future and we don’t know, I think we
shouldn’t be making a decision just yet.
And there are -- I don’t want to trash DSRC because there are advantages
of DSRC, Europe is doing a lot with it -- it’s just that when I look at the 5G versus DSRC,
DSRC is going to require a lot of government funding and spending. And, at a time when,
honestly, we don’t really do a good job of maintaining our roads, sort of the classic, I just
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
143
don’t see the political support there. I really don’t.
It could be a great technology in a laboratory, but in the real world I have
some concerns, and I think with 5G in v-to-x, you can get the private sector to pick up a lot of
those costs and it’s a benefit. I would say I am generally supportive of keeping the wireless
spectrum in the DOT.
Now, I do think at a certain point in time, we need to show in the
transportation sector that we’re really going to make good use of it because, obviously,
there's competing demands right now; but I’m not just willing to seed it to the FCC because it
hasn’t been used yet. But I do think we need to make some progress there.
MS. SNYDER: Anyone else want to weigh in on this? Okay, that was long
enough for that (laughter).
I’m curious, also, about the role of the private sector and partly in some of
the infrastructure upgrades, perhaps electrification upgrades, that need to happen to
prepare for the autonomous era.
You know, is that something that the Federal government should be doing,
or is that something that the private sector, that these companies need to do themselves i f
they want to roll out, for example, electric vehicles, that they should be in charge of figuring
out of how to also roll out electric vehicle charging stations. Whose job is that?
MS. KIGEL: Well, I don't know that it’s limiting, that it has to be the private
sector or the public sector. I think it can also be partnerships between the private sector and
public sector. It’s just like what we were talking about before; there are certain conditions
and situations for electric vehicles and electric charging stations.
If those were going to be installed and you have a significant amount of your
population dependent on that, then you have to make sure that you have a resilient
infrastructure that can accommodate that.
So, when it comes to the overall resiliency and safety, that’s usually
something that is a public sector responsibility. But I think you can work very closely in
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
144
partnership with the private sector because there are definitely some benefits. Even benefits
from the standpoint of if you're implementing infrastructure that includes sensing technology
and data collection technology and what not, that data that’s collected can be very valuable,
also, to the private sector.
And that goes to a whole other discussion about data generation, data
sharing, who owns it, that sort of thing, but I’m not so sure that that is a singular answer. I
think there can be a tremendous amount of partnership in that.
MR. MADRECKI: And I think inherently, it almost has to be a partnership.
Fast charging a Class 8 tractor trailer essentially requires the same amount of energy that’s
used in a single data power at Walmart.
So, even if you were going to have a fleet of Class A trucks, you sort of
need exponential demand that you're putting on the power grid which would seem to apply
that if like, let say my former employer UPS wanted to roll out a bunch of trucks, we might
need to coordinate with the utility and the city and others to sort of create an environment
that’s actually conducive to us doing that.
So I think there's a lot of interplay and potential to work hand in hand to
bring different technologies on line and structure rates appropriately and provide that
infrastructure where necessary.
MR. FEIGENBAUM: Yeah, in theory, the private sector should do it, but I
see nothing wrong with public/private partnerships along the lines of those just mentioned. I
think those are a win-win, and I think those are a great way to get it done.
MS. CHADHA: That’s an interesting question and I was thinking back to
two examples. One was Bird’s announcement to allocate a billion dollars last year. They
made that announcement towards the creation of bicycle lane infrastructure in cities. I’m
sure a lot of you remember that and then how, I think, six months later, if I remember
correctly, that program was closed. Not to point out Bird in particular, but I think that there is
skepticism, and maybe healthy skepticism, when private sector is pushing for some of these
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
145
infrastructure changes.
But then, on the other hand, you read cases about how some of these new
mobility companies are working with local communities that have been his torically neglected
in terms of transit investment, and how they're working at quite a grassroots level to help
bring infrastructure, whether it’s the docked bicycles, or other types of infrastructures into
those communities.
So, I think that you're right. It’s probably not a straightforward or an easy
answer. It does have to come through coordination and through partnerships.
MS. SNYDER: Talking about public/private partnerships. Ford has a
partnership with D.C. right now to introduce some autonomous vehicles. These are
happening all over the country. What will it take for some of these pilot programs to become
continuous, permanent introductions for autonomous vehicles in cities?
MR. FEIGENBAUM: I think it’s important that folks get out and experience
the technology. One of the things we've seen is if you ask a survey of people and what they
think of autonomous technology, most folks are usually negative; and they're negative
because what gets the coverage is what happened in Arizona as opposed to the number of
successful rides that another one does, because that’s really not very interesting or
newsworthy.
But we find that when folks actually ride in these vehicles and find that
they're not altogether different from today's vehicles and they're not taking them around
corners at 50 miles an hour and crashing them into other vehicles, they become much more
supportive. So, I think the pilots can sort of build on themselves, and I think it’s important
that the pilots go to different places in the city, not just one select.
I know oftentimes it can be easier just to geofence it and put it one place to
start, and I think that makes sense. And, obviously, many of these vehicles are what they
call Level 4 on the society of automotive engineer’s scale, they're not going to be
everywhere. But moving them around to different neighborhoods to increase support, I
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
146
think, is really important.
MS. KIGEL: And I’ll just add to what Baruch said. I completely agree with
that, and also to have those sustainable -- when I say sustainable, I’m not necessarily talking
about environmental. I mean sustainable from a longevity standpoint. Those sustainable
pilots are really important for information sharing. Just as I said before, there's a lot for both
parties to learn when you do that, and then the public engagement is very, very important.
Even if you look at some of our communities, those that are 65 and older,
and as you get to a point where maybe you lose your ability to be independent, and you rely
on others for mobility. When you can start to expose those members of our cities and our
states in the technology, and they can experience it, it definitely helps with public
acceptance and trust.
And that is probably one of the biggest factors that will either enhance or
impede rolling this out. So, I completely agree with what Baruch said.
MS. CHADHA: I'd add and say that in addition to these pilots, or in
conjunction with these pilots, that I’m hoping that there is robust discussion around what are
the goals that those cities are trying to achieve. So, not just simply procuring a widget, but
rather thinking about, what is the challenge, and what is the solution that they’re hoping to
experiment around.
I'd also say that it’s important to sort of think about the current mix of modes
in that city and not only where the goals of that city are, but what does this look like in a city
like D.C. where 40% of the population versus other cities in the U.S. where 10% of the
population don’t have cars.
MR. MADRECKI: Just really quickly from a pilot perspective in making it
actually last well into the future. I think we also sometimes have a tendency to want the
idealized version of autonomous vehicles to be what's piloted, rather than what is, perhaps,
very practical in the short-term and very executable on a repeating basis.
And so, if you look at things, like some of the -- and it’s not true like full AV,
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
147
of course -- but like, automation at ports; very implementable, achievable, repeatable, and
can be scaled to different places. I think that doing things like that also builds slow
acceptance over time of increasing automation. That it doesn’t just have to be like
immediately you will be able to go everywhere in an autonomous Uber.
MS. SNYDER: Jyot, I also want to open some space; NUMO works at a
global level, you just moved back from India, I would just love to hear what you're seeing
around the world.
We’ve been focusing a lot on the United States, and kind of, what you're
seeing around the places as places are beginning to wrestle with this new emerging
technology?
MS. CHADHA: Thanks, I think in the global south the rise that we are
seeing of ride-hailing has been very interesting, especially where the entire population
doesn’t have access to formalized banking products, or even sometimes internet access on
their phones. I think mobile phone penetration is quite good entities in the global south at
this point. And so, it’s interesting to see these new products that are coming out, and I'd
love to see some of those being tested over here as well for unbanked populations.
So, that’s one piece. There's, I think, a very big question around how to
regulate those models, but I think maybe the entry point of this discussion is fundamentally
different from the entry point of this discussion in the U.S., which is that the majority of trips,
for example, in a city like Bombay, are done by walking; 50% of trips is walking. Then of the
motorized, another 50% is public transport. And so, you get down to about like 8, maybe, or
10% of trips that are being made by car.
And so, in an ecosystem like that where these cities are undergoing
massive changes, new populations coming into the city, a lot of growth potential, many
aspirations for people to enter the middle class, where a vehicle has been that traditional
way to show that you have entered the middle class, I think there are very big questions
about, well, are these new models sort of substituting for people wanting to own their own
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
148
vehicle, or is it actually shifting how people think about moving around?
But also, fundamentally, the capacity of cities in the global south is a lot
lower to be able to deal with a lot of these changes. So, the hand-off of what should be
done by the public sector and the private sector, I think, is a bit more stark over there.
Aside from that slightly depressing story, I think that there also just really
exciting things happening around how we think about informal transit and power transit in
African cities, in Indian cities. Data standards that are being developed off of those. How
we’re thinking about utilizing motorcycles as taxis, as well; very large operations of those
emerging. So, there's a lot of innovation that’s taking place, but I would say just
fundamentally, it looks a bit different than what's happening in the U.S. because of that mix,
which is very different, and the affordability, which is very different.
MS. SNYDER: I’m kind of reeling from the mode-share that you just
described in India, I mean, 50% walking, 8% single-occupancy vehicle, if those even were
single-occupancy, is just so opposite from where we are. And, Baruch, you also mentioned
how autonomous technology can be helpful in transit.
I think that there's a lot of concern that autonomous technology, like TNCs,
can increase congestion because when, if taking your own door-to-door vehicle is the
cheapest, most efficient, most accessible way, then what are the incentives not to do that.
So I would love to hear more about how autonomous technology is going to get us closer to
50% closer and 8% single-occupancy vehicle, and is going to actually help the transit sector,
instead of rendering it obsolete.
MR. FEIGENBAUM: Sure, well, I don’t think we’re ever going to get to 50%
walking in those modes --
MS. SNYDER: That’s sort of a joke but --
MR. FEIGENBAUM: Maybe in New York City -- right, okay, I just didn’t
want to get anyone’s hopes up for that (laughter). So, first of all, as I think I mentioned, I
really believe pricing is the key.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
149
There's been a lot looked at urban growth boundaries, changing land use,
and all of that can work in certain situations, but what we found is that that what folks are
most sensitive to is cost. That’s what makes them change their mind. They're not, at least
by and large, sensitive to nice societal goals, as much as we might want them to. Folks
generally tend to be kind of selfish. Sorry for that downer, that’s just kind of the reali ty. So, I
think that’s important.
I would also say we may need to change our mind on what we consider
transit. Certainly, in core D.C., in core New York, the heavy rail lines are never going away.
But, in a lot of places where we have bus -- I’m a big fan of BRT; sometimes I even look at
BRT. That may not be the future.
The future may be smaller vehicles. The future may be, in fact, cars where
we actually have four people in them, as opposed to 1.1 people in them, which is basically
what we have now. And I think transit agencies, and everyone, really have to look at how
they can work together.
One of the challenges is, we've been doing transit basically the same way
for the last 50 years, and ridership is going down, and we’re not real good at change
because it’s hard. And, I’m hopeful that automated vehicles and some of these new
technologies will be the impetus for folks to actually step out of their box and we are seeing
it.
In Jacksonville, transit agencies are actually doing some interesting things,
they're testing some AVs instead of -- I think they had a people-mover system. So, some
transit agencies are looking at new technologies, which is encouraging.
MR. MADRECKI: I will add a little to that and just say that there's still only --
especially as you go into a city -- there's only so much physical real estate. And so, at some
point, you're just talking about if you can fit X number of people on a bus, X number of
people in a transit line, X number of people in an automated car, you're stil l stuck with some
of those limitations.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
150
And so, I think that that interplay of available real estate, how that real
estate is actually used and thought about as a city; how are you building this infrastructure to
facilitate cars? Are you aiming to facilitate more walking? And the leverage that your pulling
as a city plays a role in, ultimately, whether AVs cause a decline in transit use, or actually
somehow play a role in encouraging it.
MS. SNYDER: I’m going to talk about curb space too. Tom you have a
unique take on this as both an urban bicyclist and a UPS driver, and curb space is becoming
very crowded with Ubers and Lyfts, with bike lanes; there's just increasing competition and
it’s becoming more of an issue.
There are pilot programs trying to deal with this. I just kind of wanted to
hear what we've learned about how to manage curb space in a place where people are less
driving their own car and parking it, and more being picked up and let off.
MR. MADRECKI: In case this hasn’t come through yet, I tend to often take
10 steps back from questions, think about it philosophically and then go from there. So my
take on this -- and I had the luxury, and continue to have the luxury, of talking with a lot of
DOT officials all across the country, thinking about, what do you actually want to achieve,
what are your goals, how do you want your city to work and function?
And so, I straight up ask them, what are your transportation priorities if you
just listed them out. By and large, and rarely has this ever deviated, the answers I get back
are that they care most about safety, they care about equity, they care about shared
mobility, they care about basically making sure that their businesses stay in business and
they continue to have economic development, and then mention single-occupancy personal
vehicles.
And then I ask them, well then why does every street in America look the
exact opposite, right? That there's personal vehicle parking, there's probably one loading
zone; there's no bus lane, or there's some other impediment to that, I’m going to die crossing
the street, and nothing is equitable.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
151
So like, this comes -- the orientation of the curb, and when I think about
curb, it’s a political decision ultimately; and it’s demonstrating political -- well, they’ll say like,
you're curb doesn’t have to look that way, it doesn’t have to be five metered spaces that
collect a dollar an hour or something like that. You could blow up the curb entirely.
Actually, I think that D.C.’s DOT is actually very progressive in this. One of
the pilots that UPS, in conjunction with a start-up called Curb Flow, just launched actually in
D.C., is to really think about if you could take the 8 or 9 different spaces in a city that were
particularly congested, contributing to a lot of double truck parking, places like The Wharf, if
anybody’s been down there, where it’s really just not easy to get around, partially also
because of Uber and Lyft drop off, and just say, do we actually need that parking space, or
do we just turn that entire curb into a giant flex zone, essentially.
Eventually, I think the idea -- and the connection back to autonomous
vehicles here -- is that as you start to think about available curb space at designated times,
you can integrate that with whatever -- from an autonomous vehicle perspective -- so that
you know when vehicles are going to arrive, you have available curb space for them, they're
able to use that and then get on their way, and move and sort of better manage that curb
use accordingly. But I think the ultimate thing is that you don’t need to do things the way
that you do things today.
MR. FEIGENBAUM: I will say that like a lot of these issues, this is not
necessarily new, it’s just being maybe exacerbated, my automated vehicles. I mean, we
have problems today with truck loading, and bicycling, and walking and all the rest of it. I will
say -- a little plug here -- that I think 25 years ago, my boss, Adrian Moore, wrote a book
actually for Brookings that was called, Managing Curb Space, and it’s kind of cycling back
around again.
And not surprising pricing has a big part to do with it, and who you're going
to price, and how you're going to do it. And, to Tom’s point, making sure that you're having
all of the various users come in there.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
152
I think very often cities don’t necessarily consider the freight in the private
delivery. I mean, we see this a lot with designing streets. You want narrow streets because
you want to encourage certain uses, which is great; but if you're freight trucks can't get in
there and make deliveries, that’s kind of a problem.
So, making sure you have the cyclists, the walker, the freight delivery, the
people who are going to drive their cars, the Uber, Lyft, taxis, all those folks involved I think
is important.
MS. SNYDER: I want to go to the audience for questions. Does anybody --
oh, good. Yes.
SPEAKER: I wanted to take the advantage of Tom here to talk about freight
(laughter), and since you're from the Grocery Manufacturers’ Association especially,
because one of the things I learned about freight is that the actually deliver driver oftentimes
stocks the shelves. So, there's this connection that we don’t -- there's this sort of first mile,
last mile between the wholesale and the retailing freight.
So from your perspective of logistics in groceries, and that’s what you're --
what are they excited about, what are they skeptical about, what are they rolling their eyes
at, what do they think are the big problems, what can you give us as sort of insight into your
industry and this automated vehicle space?
MR. MADRECKI: Just to sort of provide some background or context, GMA
predominantly represents America’s, sort of, favorite brands that make consumer packaged
goods. It could be anybody from PepsiCo to Coca Cola, to Proctor and Gamble. So, there's
a diversity of products, not just food items, but also razor blades and soap.
That accounts for 1/5th of U.S. freight. Most of the companies in that space
rely on external service providers or are paying for transportation. They are not ultimately
the ones that have their own fleet. PepsiCo is sort of its own beast entirely and they have
own.
But, because of that there's a great deal of beholden-ness, I think, to an
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
153
industry that also -- ATA came out with their annual report yesterday talking about a truck
driver shortage that there are 60,000 fewer truck drivers today than are needed and that that
number will more than double over the next decade.
And so, I think that from our industry’s perspective, as you look to
automated vehicles, there's, I think, a hope and a promise, just from a purely efficiency
perspective from a, how do you rethink supply chains, and you're talking about a technology
that can sort of fundamentally, you know, you take something that today takes five days in a
truck because of hours of service, and things like that, and it condenses a five-day trip
across the country to two days, or even less.
So you open up this Pandora’s box of questions of like, well can you just
completely change your network configurations for supply chains, could you dramatically
change how you do business? From a driver’s side, I also don’t think that -- we talk about
autonomous vehicles -- what sometimes gets lost in that, and especially around truck driving
and the commercial side of things, is that there's this assumption that the truck driver just
goes away.
Frankly, I think that that’s missing the boat somewhat because oftentimes
that driver, and this is especially the case when I was at UPS, that driver is either the face of
your company, or they have many other roles to play. To your point about, are they stocking
shelves, or interacting in a receiving capacity, a customer service capacity; and so I think
that there's a real interest in figuring out, how do you couple that technology that can
improve safety and efficiency, within somehow enabling that driver to do more and to
provide them with additional skills and training as they go about their work.
MS. SNYDER: Other questions, yes.
SPEAKER: This is about regulation. I know FDA, for example, regulates all
the pharma companies, right. They don’t want the pharma companies to put a drug in the
market before they go through the proper trials.
I think that it was mentioned that you don’t need (inaudible) to regulate
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
154
autonomous cars. But, suppose I open a 20-person company and say, okay, I’m going to go
ahead with autonomous cars because I just buy cars from Honda, or Toyota, and add my
own lidar technology and put it in the market. Should the government allow me to do that?
MR. FEIGENBAUM: Let me just make sure I understand. Basically, what
you're saying is, should the government allow after market technologies to be placed on the
car?
SPEAKER: Or, for that matter, allow anybody to put a car out in the market
just because they can self-regulate themselves.
MR. FEIGENBAUM: I guess I would say there are certain standards that
vehicles have to submit to today in order to be certified. There is a self-certification process.
I wouldn’t say it’s true that right now you can put any technology out on the market, and I
don’t think we’d want to introduce that.
Part of me likes the entrepreneurial aspect of that, but part of me also
realizes there's a safety component as well. I think the process we have today works well.
It’s a little different than how they do it in Europe because it’s generally not super onerous for
the company, and so that allows smaller startups to do it. But there is a little -- well, there's
more than a little bit -- there's a safety component of it to make sure what's put out there isn't
going to kill people.
MS. SNYDER: Other questions? Yes, I knew if I waited long enough
somebody would have one.
SPEAKER: I’m still not sure how the technology is affecting an entire
population. They can already owe it coming with the cost. So just I’m thinking about that
kind of case; there is somebody who is making $15 per hour, and have two children, and
have to commute to somewhere, and also have to pick up child on the way the back, and
also have to go grocery shopping. For that kind of person, for that person, would benefit or
not benefit, how much she or he has to pay more or pay less?
MS. SNYDER: Are you assuming that autonomous vehicles will be
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
155
introduced as shared fleets like Uber as opposed to private car ownership?
SPEAKER: Uber or ownership or whatever. What kind of option is
available for that person? How much is that going to cost (inaudible) expenditure?
MS. SNYDER: Mm hmm. I feel that question. Busy people making
multiple trips, what's the best way to do that?
SPEAKER: [Off mic].
MS. SNYDER: I’m sorry?
SPEAKER: $15 per hour case.
MS. SNYDER: Right.
MR. FEIGENBAUM: Yeah, so that’s a tough question. But, okay, when
we’re looking at technology, at least my approach, is I’m looking if there's overall societal
gain. Are some people going to lose and some people going to win? Probably in just about
everything in life. Like, trying to make everything better for everybody is pretty challenging.
But, I do think, if you look at the shared model, which seems to be how most
people think AVs are going to roll out, at least early; and I do think, I personally do think,
there's going to be individuals owning AVs, and probably more than we think, but that
certainly does not need to be the model for everybody.
I think if you look at the shared model, because the cost of vehicle
ownership is so much less -- and the vehicles might not even be owned by folks using them;
there might actually be some sort of lease, there may be companies that are actually using
them -- I think that that is going to have a significantly lower economic cost at least over the
long run.
Now, the challenge you bring up is, how about folks that need to make trips
during the day? Are the vehicles going to be available to them, and that’s challenging.
That’s one of the questions folks are looking at with these shared vehicles; if everyone
leaves work at 5 o’clock, are there going to be enough vehicles and how do you guarantee
that. And, I honestly think that’s a little bit tricky. I think that one of the reasons individually
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
156
owned vehicles will be available.
But I do think there is a lot more potential, and there's also -- we have
programs right now that are used if folks, for example, take transit one way but the transit
doesn’t operate back; or they're in van pools, which are where 7 to 14 people share a ride;
things called Guaranteed Ride Home, where there's a way to get home or get some place if
they need to.
So I think we can expand those types of things. I think we can make it work.
I’m not saying we've got it solved right now, but I think it’s a very promising technology. And
I do think it can make driving more affordable for folks.
MS. SNYDER: People are --
MS. CHADHA: Can I quickly add something to that?
MS. SNYDER: Sure, of course.
MS. CHADHA: I think your question is prompting me to think about an
inclusive design process. And I think the private sector is much better at this sometimes.
But when you look at how a lot of transit options, or how cities are planned, I think the trips
that you're mentioning are not at the forefront for planning those systems.
And so, I very much agree with you that that design process needs to be
much more inclusive if we’re really going to push for transport that is accessible and useful
for all.
MS. SNYDER: I think that that gets to the point also of, what if we put
equity first, and framed all of this as we’re trying to solve a problem of equity. We can on top
of that, also, congestion. We can add on top of it, mobility. And then, what is the best way
for that parent working a $15 an hour job to get around.
I also wanted to get back to Baruch’s point about everyone leaving at 5
o’clock, will there be enough vehicles? I mean, that’s a nightmare now. That is a failure of
our transportation system, and a failure of our ability to work around it. If people just had
more staggered work times, so many of our transportation problems could be solved.
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
157
MR. FEIGENBAUM: I would just say -- and we've tried things. Obviously,
we've tried to incentivize carpooling, to get folks to carpool. And it’s a little better in the D.C.
region because the congestion is so bad, but by and large, we've not been super successful.
I, as someone who gets to work from home a lot, don’t understand why folks
aren’t willing to go in an hour earlier or stay an hour later to avoid the peak congestion. And
I know it’s sometimes not that easy, you’ve got to pick a child up from daycare, whatever.
But, again, not to be a broken record, but I think looking at pricing -- taking
the I66 lanes -- for those of you who are a little unfamiliar with that. Now, the ones in the
morning are a bit of a challenge and it has to do with some geometric design issues. We’ve
got to merge from four lanes to two lanes and also getting into D.C., so okay, but the ones in
the afternoon -- I still support them by the way, I’ll be happy to talk to you about why -- the
ones in the afternoon, though, are really wonderful.
If you look at it, a relatively minor price of maybe 50 to 75 cents per mile,
has basically gotten rid of all of the congestion on I66 westbound, going from D.C. out to
Fairfax County where I live, and it’s actually kind of amazing because it’s really not that
much in the context of D.C. And, honestly, a lot of the incomes here are fairly high, not all of
them, but some of them.
And the biggest problem is what happens when the pricing ends and you
have congestion before and after because everyone is trying to avoid paying the price,
which I totally get, but if you priced it 24/7 you would not have this problem (laughter).
So, there's a lot we can do with policy. I realize the political acceptance is
very, very challenging, but to your point, we can solve these problems if we really want to, I
think.
MS. SNYDER: Any other questions from the audience? Yes.
SPEAKER: Thank you, hi, I’m an urban planning student interning at the
Federal Highway Administration, who comes from a really rural part of the country, and while
we have talked today a lot about these issues in urban places, it has also been alluded to
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
158
that this, too, may change the form in rural areas. And so, I’m just curious, from your broad
range of expertise and perspectives here, how you guys see this playing out in more rural
areas in the U.S.?
MR. FEIGENBAUM: It’s going to be different. And I guess, one of the
questions in my mind is, will there be enough folks going to similar locations where the
ridesharing is going to work? Or is it going to be more individually owned vehicles?
I mean, it’s definitely going to be more individually owned, that’s just reality.
And, obviously, there's been some talk of electrification, and I do think Tesla says a lot of
stuff, but they're success with the model 3, in terms of the demand and also producing it,
makes me fairly optimistic for electrification. So, I think that’s positive there, assuming the
power source, of course.
We don’t know is the question. We think for the exurban areas there's going
to be quite a bit of ridesharing. I think for the rural areas, I don't know. Maybe some of the
new models can work to improve the transit service in the rural areas, which is, you know,
basically nonexistent, or could actually do a lot with Dial-a-Ride sort of type, but it’s going to
be very different.
I think you bring up a good point and I think it’s going to come to the urban
areas first before it gets to the rural areas, but that’s obviously a politically sensitive topic.
MS. SNYDER: With the few seconds that we have left, I just want to ask
what your biggest worries are?
I don't know if this is a few-second answer. But, you know, as we approach
right now, with very little regulation, as we approach the advent of the driverless era, what
keeps you up at night?
MS. CHADHA: That after all of this, that we still end up with a transportation
system that is the second highest household expense, that results in many deaths, that is a
big contributor to emissions.
MR. FEIGENBAUM: I have a two-part answer. Not surprisingly, I’m worried
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
159
more about over regulation, certainly not right now, but in the future. And, I also worry that
we will not take this opportunity to do things such as congestion pricing that could be a big
benefit.
MS. KIGEL: One of the things that I do think about has to do with funding.
Funding overall transportation infrastructure, especially when we’re going with the
electrification of vehicles. And I think we have a great opportunity here because we’re
talking about a lot of innovation. And as we’re talking about innovation, and also as we
might be moving towards shared mobility and whatnot, we should really look innovatively as
to how we actually fund infrastructure.
It’s been challenging, particularly, because we have states that do their own
level of funding, and then we have the Federal government that’s also responsible for
funding. So, as states are experimenting in different strategies to be innovative, we also
have the Federal government who covers this fear of the entire country, and we could have
the ability of different methods being used in different places, and that could create
challenges for us.
MR. MADRECKI: I wish I could tell you that political decisions were made
just on the basis of fact (laughter), and like complete logic. So, I think that my fear is political
realities, and sort of, the desire of people to appease other people gets in the way of some
of the decisions that are crucially needed.
It could be to Baruch’s point about congestion pricing, or some other
mechanism of changing transportation. But it can also be about just the way that things are
rolled out.
And so, if it’s more politically viable in the short-term to have a personal
vehicle regulatory approach for autonomous vehicles rather than commercial applications --
which I actually, and this may be -- Yes, I represent GMA, but I think that there's more viable
short-term options for commercial autonomy than there are for personal autonomy. But
that’s not a politically viable thing, necessarily, to say. And so, I think the sensitivity there
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
160
would have to be just that you don’t let politics get in the way of actual progress.
MS. SNYDER: Well that does it for us. Please join me in thanking the
panel for this wonderful conversation (Applause).
* * * * *
CARS-2019/07/25
ANDERSON COURT REPORTING
1800 Diagonal Road, Suite 600
Alexandria, VA 22314
Phone (703) 519-7180 Fax (703) 519-7190
161
CERTIFICATE OF NOTARY PUBLIC
I, Carleton J. Anderson, III do hereby certify that the forgoing electronic file
when originally transmitted was reduced to text at my direction; that said transcript is a true
record of the proceedings therein referenced; that I am neither counsel for, related to, nor
employed by any of the parties to the action in which these proceedings were taken; and,
furthermore, that I am neither a relative or employee of any attorney or counsel employed by
the parties hereto, nor financially or otherwise interested in the outcome of this action.
Carleton J. Anderson, III
(Signature and Seal on File)
Notary Public in and for the Commonwealth of Virginia
Commission No. 351998
Expires: November 30, 2020